JP6935553B2 - Methods and Devices for Coding / Decoding Motion Vectors Based on Reduced Predicted Motion Vector Candidates - Google Patents

Description

The present invention relates to a method and an apparatus for encoding / decoding a motion vector, and more specifically, to a method and an apparatus for predictively coding and predictively decoding a motion vector of a current block.

MPEG (moving picture experts group) -4H. Codecs such as 264 / MPEG-4 AVC (advanced video coding) utilize the motion vector of a previously encoded block adjacent to the current block to predict the motion vector of the current block. The median of the motion vector of the previously encoded block adjacent to the left, top and top right of the current block is used as the motion vector predictor of the current block. Instead of encoding the motion vector of the current block as it is, only the motion vector difference between the motion vector of the current block and the predicted motion vector is encoded.

One or more exemplary embodiments of the invention provide a method and apparatus for predictively coding and predictively decoding motion vectors, providing a computer-readable recording medium on which a program for performing the method is recorded. It is to provide.

The method of encoding the motion vector according to the embodiment of the present invention for solving the technical problem estimates the motion vector of the current block, and based on the estimation result, among all the candidates of the predicted motion vector. A step of determining a first predicted motion vector candidate as a predicted motion vector of the current block and generating information about the motion vector based on the motion vector of the current block and the predicted motion vector of the current block. Among the overall candidates, a virtual motion vector is generated using the second predicted motion vector candidate and the difference vector, a difference vector between the virtual motion vector and each of the overall candidates is generated, and the difference vector is generated. And the information about the motion vector, and the step of selectively excluding the second predicted motion vector candidate from the overall candidate, the information about the motion vector, and the predicted motion vector of the current block. Includes a step of encoding information.

The method for decoding the motion vector according to the embodiment of the present invention for solving the technical problem is predetermined among the stage of decoding the information about the motion vector of the current block and the overall candidate of the predicted motion vector. A virtual motion vector is generated by using the predicted motion vector candidate of the above and the information about the decoded motion vector, a difference vector between the virtual motion vector and each of the overall candidates is generated, and the generation is performed. The difference vector is compared with the information about the decoded motion vector, and the predetermined predicted motion vector is selectively excluded from the overall candidates, and the candidate of the predicted motion vector that is not excluded. One of them is determined as the predicted motion vector of the current block, and the motion vector of the current block is restored based on the determined predicted motion vector and the information about the decoded motion vector. ,including.

An apparatus for encoding a motion vector according to an embodiment of the present invention for solving the technical problem estimates a motion vector of a current block, and based on the estimation result, among all candidates of a predicted motion vector. A motion vector estimation that determines a first predicted motion vector candidate as a predicted motion vector of the current block and generates information about the motion vector based on the motion vector of the current block and the predicted motion vector of the current block. Part; Among the overall candidates, a virtual motion vector is generated using the second predicted motion vector candidate and the difference vector, and a difference vector between the virtual motion vector and each of the overall candidates is generated, and the above. A candidate determination unit that compares information about the difference vector and the motion vector and selectively excludes the second predicted motion vector candidate from the overall candidates; and information about the motion vector and prediction of the current block. Includes a motion vector estimator; which encodes information about the motion vector.

The device for decoding the motion vector according to the embodiment of the present invention for solving the above technical problem is a motion vector decoding unit that decodes information about the motion vector of the current block; Among them, a virtual motion vector is generated by using information about a predetermined predicted motion vector candidate and the decoded motion vector, and a difference vector between the virtual motion vector and each of the overall candidates is generated. , A candidate determination unit that compares the generated difference vector with information about the decoded motion vector and selectively excludes the predetermined predicted motion vector from the overall candidates; and a non-excluded prediction. One of the motion vector candidates is determined as the predicted motion vector of the current block, and the motion vector of the current block is based on the information about the determined predicted motion vector and the decoded motion vector. Includes a motion vector restorer that restores.

To solve the technical problems, the present invention provides a computer-readable recording medium on which a program for performing a method of encoding and / or decoding the motion vector is recorded.

According to the present invention, even when the motion vector is predicted-encoded and predicted-decoded by using the predicted motion vector candidates, the number of predicted motion vector candidates is reduced and the motion vector is predicted-encoded and predicted-decoded. can do. Therefore, among the candidates for the predicted motion vector, the information required to identify the predicted motion vector candidate used for the prediction of the motion vector of the current block can be encoded with the minimum number of bits, and the motion vector coding can be performed. / Decoding compression ratio is improved, which can also improve video coding / decoding compression ratio.

It is a drawing which illustrates the image coding apparatus by one Embodiment of this invention. It is a drawing which illustrates the image decoding apparatus by one Embodiment of this invention. It is a drawing which illustrates the hierarchical coding unit by one Embodiment of this invention. It is a drawing which illustrates the image coding part based on the coding unit by one Embodiment of this invention. It is a drawing which illustrates the image decoding part based on the coding unit by one Embodiment of this invention. It is a drawing which illustrates the maximum coding unit, the sub-coding unit and the prediction unit by one Embodiment of this invention. It is a drawing which illustrates the coding unit and conversion unit by one Embodiment of this invention. It is a drawing which illustrates the division | division form of the coding unit, the prediction unit and the conversion unit by one Embodiment of this invention. It is a drawing which illustrates the division | division form of the coding unit, the prediction unit and the conversion unit by one Embodiment of this invention. It is a drawing which illustrates the division | division form of the coding unit, the prediction unit and the conversion unit by one Embodiment of this invention. It is a drawing which illustrates the division | division form of the coding unit, the prediction unit and the conversion unit by one Embodiment of this invention. It is a drawing which illustrates the apparatus which encodes the motion vector by one Embodiment of this invention. It is a drawing which illustrates the candidate of the predicted motion vector by one Embodiment of this invention. It is a drawing which illustrates the candidate of the predicted motion vector by one Embodiment of this invention. It is a drawing which illustrates the block of various sizes adjacent to the present block by one Embodiment of this invention. It is a drawing which illustrates the block of various sizes adjacent to the present block by one Embodiment of this invention. It is a drawing which illustrates the block of various sizes adjacent to the present block by one Embodiment of this invention. It is a drawing which illustrates the candidate of the predicted motion vector by another embodiment of this invention. It is a drawing which illustrates the candidate of the predicted motion vector by another embodiment of this invention. It is a drawing which illustrates the candidate of the predicted motion vector by another embodiment of this invention. It is a drawing which illustrates the method of reducing the candidate of the predicted motion vector by one Embodiment of this invention. It is a drawing which illustrates the position of the present block included in the coding unit of a predetermined size by one Embodiment of this invention. It is a drawing which illustrates the position of the present block included in the coding unit of a predetermined size by one Embodiment of this invention. It is a drawing which illustrates the position of the present block included in the coding unit of a predetermined size by one Embodiment of this invention. It is a drawing which illustrates the position of the present block included in the coding unit of a predetermined size by one Embodiment of this invention. It is a drawing which illustrates the apparatus which decodes the motion vector by one Embodiment of this invention. It is a flowchart for demonstrating the method of coding a motion vector by one Embodiment of this invention. It is a flowchart for demonstrating the method of decoding the motion vector by one Embodiment of this invention.

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. An expression such as "at least one ~" modifies the entire list of components and does not change the individual components of the list if it precedes the list of components.

Hereinafter, the video indicates a still video of the video, or the video and the video itself.

FIG. 1 illustrates a video coding device according to an embodiment of the present invention. Referring to FIG. 1, the video coding apparatus 100 according to the embodiment of the present invention includes a maximum coding unit division unit 110, a coding depth determination unit 120, a video data coding unit 130, and a coding information coding unit 140. including.

The maximum coding unit division unit 110 can divide the current frame or the current slice based on the maximum coding unit, which is the maximum size coding unit. The current frame or current slice can be divided into at least one maximum coding unit.

According to one embodiment of the present invention, the coding unit is also represented by utilizing the maximum coding unit and the depth. As described above, the maximum coding unit indicates the coding unit having the largest size among the coding units of the current frame, and the depth indicates the degree to which the coding units are hierarchically reduced. As the depth increases, the coding unit is reduced from the maximum coding unit to the minimum coding unit, the depth of the maximum coding unit is defined as the minimum depth, and the depth of the minimum coding unit is defined as the maximum depth. It will also be done. As the maximum coding unit increases in depth, the size of the depth-based coding unit decreases, and the k-depth sub-coding unit may include a plurality of sub-coding units having a depth greater than k.

As the size of the encoded frame increases, if the image is encoded in a larger unit, the image can be encoded with a higher image compression rate. However, if the coding unit is increased and the size is fixed, it is not possible to efficiently encode the image by reflecting the ever-changing characteristics of the image.

For example, when coding a flat area related to the sea or the sky, the larger the coding unit, the better the compression ratio, but when coding a complex area related to a person or a building, the smaller the coding unit. The more the compression ratio is improved.

To this end, one embodiment of the present invention sets a different size maximum video coding unit for each frame or slice and sets the maximum depth. Since the maximum depth means the maximum number of times that the coding unit is reduced, the minimum coding unit size included in the maximum video coding unit can be variably set by the maximum depth.

The coding depth determination unit 120 determines the maximum depth. The maximum depth is determined based on the rate-distortion cost calculation. The maximum depth may be determined differently for each frame or slice, or differently for each maximum coding unit. The determined maximum depth is output to the coding information coding unit 140, and the video data for each maximum coding unit is output to the video data coding unit 130.

The maximum depth means the smallest size coding unit included in the maximum coding unit, that is, the minimum coding unit. In other words, the maximum coding unit is also divided into sub-coding units of different sizes at different depths. Details will be described later with reference to FIGS. 8A to 8D. Further, the sub-coding units of different sizes included in the maximum coding unit are predicted or converted based on the processing units of different sizes. The transformation transforms the pixel values of the spatial domain into the coefficients of the frequency domain, which may be the discrete cosine transform or the KLT (Karhunen Loever transform).

In other words, the video coding apparatus 100 can perform a plurality of processing steps for video coding based on processing units of various sizes and various forms. For coding video data, processing steps such as prediction, conversion, and entropy coding are performed, but processing units of the same size are used in all stages, and processing units of different sizes are used for each stage. You can also do it.

For example, the video coding apparatus 100 can select a processing unit different from the coding unit in order to predict a predetermined coding unit.

When the size of the coding unit is 2Nx2N (where N is a positive constant), the processing unit for prediction is 2Nx2N, 2NxN, Nx2N, NxN, or the like. In other words, motion prediction may be performed based on a processing unit in the form of halving at least one of the height or width of the coding unit. Hereinafter, the processing unit on which the prediction is based will be referred to as a “prediction unit”.

The prediction mode is at least one of an intra mode, an inter mode and a skip mode, and the specific prediction mode may also be performed only for a prediction unit of a particular size or form. For example, the intra mode is performed only for predictive units of square 2Nx2N, NxN size. Further, the skip mode is executed only for the prediction unit of 2Nx2N size. If there are a plurality of prediction units inside the coding unit, prediction is performed for each prediction unit, and the prediction mode with the smallest coding error is also selected.

Further, the video coding apparatus 100 can convert video data based on a processing unit having a size different from that of the coding unit. For the conversion of the coding unit, the conversion is also performed based on the data unit of the size smaller than or the same as the coding unit. Hereinafter, the processing unit that is the basis of conversion will be referred to as a “conversion unit”.

The coding depth determination unit 120 can determine the sub-coding unit included in the maximum coding unit by using the rate-distortion optimization based on the Lagrangian multiplier. .. In other words, it is possible to determine what form the maximum coding unit is divided into a plurality of sub-coding units, where the plurality of sub-coding units vary in size depending on the depth. After that, the video data coding unit 130 encodes the maximum coding unit and outputs a bit stream based on the division form determined by the coding depth determination unit 120.

Coding information The coding unit 140 encodes information about the coding mode of the maximum coding unit determined by the coding depth determination unit 120. It encodes information about the division form of the maximum coding unit, information about the maximum depth, and information about the coding mode of the sub-coding unit for each depth, and outputs a bit stream. The information about the coding mode of the sub-coding unit may include information about the prediction unit of the sub-coding unit, prediction mode information for each prediction unit, information about the conversion unit of the sub-coding unit, and the like.

The information about the division form of the maximum coding unit may be information indicating the division of each coding unit. For example, when the maximum coding unit is divided and coded, the information indicating the division is encoded with respect to the maximum coding unit, and the sub-coding unit generated by dividing the maximum coding unit is further divided. Also in the case of encoding, the information indicating the division is encoded for each sub-coding unit. The information indicating the division can be the flag information indicating the division.

Since there are sub-coding units of different sizes for each maximum coding unit and information about the coding mode must be determined for each sub-coding unit, at least one for one maximum coding unit. Information about one coding mode is determined.

The video coding apparatus 100 can generate a sub-coding unit by halving the height and width of the maximum coding unit as the depth increases. That is, if the size of the coding unit of k depth is 2Nx2N, the size of the coding unit of k + 1 depth is NxN.

Therefore, the video coding apparatus 100 according to one embodiment determines the optimum division form for each maximum coding unit based on the size and maximum depth of the maximum coding unit in consideration of the characteristics of the video. Can be done. By variably adjusting the size of the maximum coding unit in consideration of the image characteristics and dividing the maximum coding unit into sub-coding units of different depths to encode the image, images of various resolutions can be obtained. Can be encoded more efficiently.

FIG. 2 illustrates a video decoding apparatus according to an embodiment of the present invention. Referring to FIG. 2, the video decoding apparatus 200 according to the embodiment of the present invention includes a video data acquisition unit 210, a coded information extraction unit 220, and a video data decoding unit 230.

The video-related data acquisition unit 210 parses the bit stream received by the video decoding device 200, acquires video data for each maximum coding unit, and outputs the video data to the video data decoding unit 230. The video data acquisition unit 210 can extract information about the maximum coding unit of the current frame or slice from the header related to the current frame or slice. In other words, the bit stream is divided into the maximum coding units, and the video data decoding unit 230 is made to decode the video data for each maximum coding unit.

The coding information extraction unit 220 parsing the bit string received by the video decoding apparatus 200, and from the header related to the current frame, the maximum coding unit, the maximum depth, the division form of the maximum coding unit, and the code of the sub-coding unit. Extract information about the conversion mode. Information about the division mode and the coding mode is output to the video data decoding unit 230.

The information about the division form of the maximum coding unit may include information about sub-coding units of different sizes depending on the depth contained in the maximum coding unit. As described above, the information about the division form may be information (for example, flag information) indicating the division that is encoded for each coding unit. The information about the coding mode may include information about the prediction unit for each sub-coding unit, information about the prediction mode, information about the conversion unit, and the like.

The video data decoding unit 230 decodes the video data of each maximum coding unit based on the information extracted by the coded information extraction unit, and restores the current frame.

Based on the information about the division form of the maximum coding unit, the video data decoding unit 230 can decode the sub-coding unit included in the maximum coding unit. The decoding process may include an inter-prediction process including intra-prediction and motion compensation and an inverse transformation process.

The video data decoding unit 230 can perform intra-prediction or inter-prediction based on the information about the prediction unit for each sub-coding unit and the information about the prediction mode for the prediction of the sub-coding unit. Further, the video data decoding unit 230 can perform inverse transformation for each sub-coding unit based on the information about the conversion unit of the sub-coding unit.

FIG. 3 illustrates a hierarchical coding unit according to an embodiment of the present invention. Referring to FIG. 3, the hierarchical coding unit according to the invention may include 32x32, 16x16, 8x8 and 4x4 from a coding unit having a width x height of 64x64. In addition to the square coding unit, there can be coding units having width x height of 64x32, 32x64, 32x16, 16x32, 16x8, 8x16, 8x4, 4x8.

Referring to FIG. 3, the maximum coding unit size is set to 64x64 and the maximum depth is set to 2 for the video data 310 having a resolution of 1920x1080.

For the video data 320 having another resolution of 1920x1080, the size of the maximum coding unit is set to 64x64 and the maximum depth is set to 4. For the video data 330 having a resolution of 352x288, the size of the maximum coding unit is set to 16x16 and the maximum depth is set to 2.

When the resolution is high or the amount of data is large, it is desirable that the maximum size of the coding size is relatively large in order not only to improve the compression rate but also to accurately reflect the video characteristics. Therefore, in the video data 310 and 320 having a higher resolution than the video data 330, the size of the maximum coding unit is selected to be 64x64.

Maximum depth indicates the total number of layers in the hierarchical coding unit. Since the maximum depth of the video data 310 is 2, the coding unit 315 of the video data 310 has a major axis size of 32, 16 as the depth increases from the maximum coding unit having a major axis size of 64. It may include sub-coding units.

On the other hand, since the maximum depth of the video data 330 is 2, the coding unit 335 of the video data 330 has a major axis size of 8 or 4 as the depth increases from the maximum coding unit having a major axis size of 16. It may include up to the coding unit which is.

Since the maximum depth of the video data 320 is 4, the coding unit 325 of the video data 320 has a major axis size of 32, 16, 8 as the depth increases from the maximum coding unit having a major axis size of 64. It may include up to a sub-coding unit of 4. As the depth increases, the image is encoded based on a smaller sub-coding unit, which makes it suitable for encoding an image containing more detailed scenes.

FIG. 4 illustrates a video coding unit based on a coding unit according to an embodiment of the present invention.

In the current frame 405, the intra prediction unit 410 makes an intra prediction for the prediction unit in the intra mode, and the motion estimation unit 420 and the motion compensation unit 425 make the current frame 405 and the reference frame for the prediction unit in the inter mode. Inter-prediction and motion compensation are performed using 495.

Residual values are generated based on the prediction units output from the intra prediction unit 410, motion estimation unit 420, and motion compensation unit 425, and the generated residual values are the conversion unit 430 and the quantization unit 440. Is output as a quantized conversion coefficient.

The quantized conversion coefficient is restored to the residual value again via the inverse quantization unit 460 and the inverse conversion unit 470, and the restored residual value is post-processed through the deblocking unit 480 and the loop filtering 490. Is output to the reference frame 495. The quantized conversion coefficient is output to the bit stream 455 via the entropy coding unit 450.

In order to encode by the video coding method according to the embodiment of the present invention, the intra prediction unit 410, the motion estimation unit 420, the motion compensation unit 425, the conversion unit 430, and the quantization unit, which are the components of the video coding unit 400, are used. The 440, the entropy coding unit 450, the inverse quantization unit 460, the inverse conversion unit 470, the deblocking unit 480, and the loop filtering 490 are all set to the maximum coding unit, the sub-coding unit by depth, the prediction unit, and the conversion unit. Based on this, the video coding process is processed.

FIG. 5 illustrates a video decoding unit based on a coding unit according to an embodiment of the present invention.

The bit stream 505 passes through the parsing unit 510, and the encoded video data to be decoded and the coding information necessary for decoding are parsed. The encoded video data is output as dequantized data via the entropy decoding unit 520 and the dequantization unit 530, and is restored to the residual value via the inverse conversion unit 540. The residual value is added to the result of the intra prediction of the intra prediction unit 550 or the motion compensation result of the motion compensation unit 560, and is restored for each coding unit. The restored coding unit is used for predicting the next coding unit or the next frame via the deblocking unit 570 and the loop filtering 580.

In order to decode by the video decoding method according to the embodiment of the present invention, the parsing unit 510, the entropy decoding unit 520, the inverse quantization unit 530, the inverse conversion unit 540, and the intra, which are the components of the video decoding unit 400, are used. The prediction unit 550, the motion compensation unit 560, the deblocking unit 570, and the loop filtering 580 all process the video decoding process based on the maximum coding unit, the sub-coding unit by depth, the prediction unit, and the conversion unit. ..

In particular, the intra prediction unit 550 and the motion compensation unit 560 determine the prediction unit and the prediction mode in the sub-coding unit in consideration of the maximum coding unit and the depth, and the inverse conversion unit 540 determines the size of the conversion unit. Consider and perform the inverse transformation.

FIG. 6 illustrates the maximum coding unit, sub-coding unit and prediction unit according to one embodiment of the present invention.

The video coding device 100 and the video decoding device 200 according to the embodiment of the present invention utilize hierarchical coding units in order to perform coding / decoding in consideration of video characteristics. The maximum coding unit and the maximum depth may be set adaptively according to the characteristics of the image, or may be variously set according to the user's request.

The hierarchical structure 600 of the coding unit according to one embodiment of the present invention illustrates the case where the maximum coding unit 610 has a height and a width of 64 and a maximum depth of 4. The depth increases along the vertical axis of the hierarchical structure 600 of the coding units, and the increase in depth reduces the width and height of the sub-coding units 620 to 650. Further, along the horizontal axis of the hierarchical structure 600 of the coding units, the prediction units of the maximum coding unit 610 and the sub-coding units 620 to 650 are shown.

The maximum coding unit 610 has a depth of 0 and a size of the coding unit, i.e., width and height of 64x64. Depth increases along the vertical axis, size 32x32 depth 1 sub-coding unit 620, size 16x16 depth 2 sub-coding unit 630, size 8x8 depth 3 sub-coding unit 640, size. There is a 4x4 depth 4 subcoding unit 650. The sub-coding unit 650 having a depth of 4 having a size of 4x4 is the minimum coding unit.

With reference to FIG. 6, examples of prediction units are illustrated along the horizontal axis for each depth. That is, the prediction unit of the maximum coding unit 610 at depth 0 is the same as or smaller than the coding unit 610 of size 64x64, the prediction unit 610 of size 64x64, the prediction unit 612 of size 64x32, and the size. The 32x64 prediction unit 614 and the size 32x32 prediction unit (616 may be used).

The prediction unit of the size 32x32 coding unit 620 at depth 1 is the same as or smaller than the size 32x32 coding unit 620, the size 32x32 prediction unit 620, the size 32x16 prediction unit 622, and the size. It may be a 16x32 prediction unit 624 and a size 16x16 prediction unit 626.

The prediction unit of the size 16x16 coding unit 630 at depth 2 is the same as or smaller than the size 16x16 coding unit 630, the size 16x16 prediction unit 630, the size 16x8 prediction unit 632, and the size. It may be an 8x16 prediction unit 634 and a size 8x8 prediction unit 636.

The prediction unit of the size 8x8 coding unit 640 at depth 3 is the same as or smaller than the size 8x8 coding unit 640, the size 8x8 prediction unit 640, the size 8x4 prediction unit 642, and the size. It may be a 4x8 prediction unit 644 and a size 4x4 prediction unit 646.

Finally, the size 4x4 coding unit 650 of depth 4 is the maximum depth coding unit and the prediction unit is the size 4x4 prediction unit 650. However, as the maximum depth coding unit, the size of the coding unit and the prediction unit do not necessarily have to be the same, and like the other coding units 610 to 650, the size is smaller than the coding unit. It is also possible to make a prediction by dividing it into the prediction units of.

FIG. 7 illustrates a coding unit and a conversion unit according to an embodiment of the present invention.

The video coding apparatus 100 and the video decoding apparatus 200 according to the embodiment of the present invention encode the maximum coding unit as it is, or the maximum coding into sub-coding units that are smaller than or the same as the maximum coding unit. The unit is divided and encoded. During the coding process, the size of the conversion unit for conversion is also selected to be the size for the highest compression ratio, regardless of the coding unit and the prediction unit. For example, when the coding unit 710 is currently 64x64 size, conversion is also performed using the conversion unit 720 of 32x32 size.

8A to 8D illustrate the division of the coding unit, the prediction unit, and the conversion unit according to the embodiment of the present invention.
8A and 8B illustrate coding units and prediction units according to one embodiment of the invention.

FIG. 8A illustrates a division mode selected by the video coding apparatus 100 according to an embodiment of the present invention to encode the maximum coding unit 810. The video coding apparatus 100 divides the maximum coding unit 810 into various forms, encodes them, and then compares the coding results of the various divided forms based on the RD cost to determine the optimum divided form. select. When it is optimal to encode the maximum coding unit 810 as it is, the maximum coding unit 800 may be encoded without dividing the maximum coding unit 810 as shown in FIGS. 8A to 8D. can.

Referring to FIG. 8A, the maximum coding unit 810 having a depth of 0 is divided into sub-coding units having a depth of 1 or more and coded. After the maximum coding unit 810 is divided into four depth 1 sub-coding units, all or part of the depth 1 sub-coding units are further divided into depth 2 sub-coding units.

Of the sub-coding units at depth 1, the sub-coding units located at the upper right side and the sub-coding units located at the lower left side were divided into sub-coding units having a depth of 2 or more. A part of the sub-coding units having a depth of 2 or more is further divided into sub-coding units having a depth of 3 or more.

FIG. 8B illustrates the division form of the prediction unit related to the maximum coding unit 810. Referring to FIG. 8B, the prediction unit 860 related to the maximum coding unit is divided differently from the maximum coding unit 810. In other words, the prediction unit for each sub-coding unit is smaller than the sub-coding unit.

For example, among the sub-coding units having a depth of 1, the prediction unit related to the sub-coding unit 854 located at the lower right side may be smaller than the sub-coding unit 854. Of the sub-coding units 814,816,818,828,850,852 at depth 2, the prediction unit related to some of the sub-coding units 815,816,850,852 may be smaller than the sub-coding units.

Further, the prediction unit related to the sub-coding unit 822, 832, 848 at the depth 3 may be smaller than the sub-coding unit. The prediction unit is a form in which each sub-coding unit is halved in the height or width direction, or is divided into four in the height and width direction.

8C and 8D illustrate predictive units and conversion units according to one embodiment of the present invention.

FIG. 8C illustrates the division form of the prediction unit related to the maximum coding unit 810 shown in FIG. 8B, and FIG. 8D illustrates the division form of the conversion unit of the maximum coding unit 810.

With reference to FIG. 8D, the division form of the conversion unit 870 may be set differently from the prediction unit 860.

For example, even if the prediction unit related to the coding unit 854 at depth 1 is selected in a form in which the height is halved, the conversion unit is also selected to have a size equal to the size of the coding unit 854 at depth 1. .. Similarly, even if the prediction unit related to the coding unit 814,850 at depth 2 is selected in a form in which the height of the coding unit 814,850 at depth 2 is halved, the conversion unit is the coding at depth 2. It is also selected to be equal in size to the original size of the units 814 and 850.

The conversion unit may be selected to a size smaller than the prediction unit. For example, when the prediction unit related to the coding unit 852 of depth 2 is selected in a form in which the width is halved, the conversion unit is a form in which the height and width are halved, which is a size smaller than the prediction unit. Also selected for.

FIG. 9 illustrates an apparatus that encodes a motion vector according to an embodiment of the present invention.

A device included in the video coding device 100 of FIG. 1 or the video coding unit 400 of FIG. 4 and encoding a motion vector is illustrated in detail in FIG. Referring to FIG. 9, the motion vector coding apparatus 900 according to the embodiment of the present invention includes a motion vector estimation unit 910, a candidate determination unit 920, and a motion vector coding unit 930.

In order to decode an encoded block using inter-prediction, or temporal prediction, information about a motion vector that indicates the relative position difference between the current block and a similar block in the reference picture. is required. Therefore, at the time of video coding, the information about the motion vector is encoded and inserted into the bit stream, but if the information about the motion vector is encoded and inserted as it is, the overhead for encoding the information about the motion vector ( overhead) increases and the compression rate of video data decreases.

Therefore, in video coding, the motion vector of the current block is predicted, and only the motion vector difference between the predicted motion vector (motion vector predictor) generated as a result of the prediction and the original motion vector is encoded. By inserting it into the bit stream, it also compresses the information about the motion vector.

The predictive coding of the motion vector may have an explicit mode and an implicit mode.

MPEG-4 H. Codecs such as 264 / MPEG-4 AVC (advanced video coding) utilize the motion vector of a previously encoded block adjacent to the current block to predict the motion vector of the current block. The median of the motion vector of the previously encoded block adjacent to the left, top and top right of the current block is used as the predicted motion vector of the current block. Since the motion vectors of all the blocks encoded by using the inter-prediction are predicted by using the same method, the information related to the predicted motion vector of the current block does not need to be encoded separately. However, the video coding apparatus 100 or the video coding unit 400 according to the present invention does not separately encode the above-mentioned information about the predicted motion vector in order to predict the motion vector more accurately. And use explicit modes to encode information about predicted motion vectors. Explicit mode means a mode that encodes information about the predicted motion vector used for the predicted motion vector of the current block among multiple predicted motion vector candidates and inserts it into the bitstream as a sequence parameter, slice parameter, or block parameter. do.

FIG. 9 illustrates a device that performs predictive coding when a motion vector is coded in such an explicit mode. The motion vector estimation unit 910 estimates the motion vector of the current block. A block similar to or the same as the current block is searched for at least one reference picture, and a motion vector which is a relative positional difference between the current block and the searched reference block is estimated based on the search result. A block similar to or the same as the current block can be searched based on the calculation of SAD (sum of absolute difference), and the motion vector of the current block can be estimated based on the search result.

The motion vector estimation unit 910 also predicts the motion vector of the current block based on the motion vector of the block contained in the previously encoded region adjacent to the current block. In other words, the motion vectors of the blocks contained in the previously encoded region adjacent to the current block are set as predicted motion vector candidates (candidates), and the estimated current block among the predicted motion vector candidates is set. Determine the predicted motion vector candidates that are most similar to the motion vector of.

MPEG-4 H. Codecs such as 264 / MPEG-4 AVC utilize the median motion vector of a previously encoded block adjacent to the current block on the left, top and top right as the predicted motion vector of the current block. Since all encoded blocks are predicted using the motion vector of the previously encoded block and only one predicted motion vector is used, the information related to the predicted motion vector needs to be encoded separately. There is no. In other words, the predicted motion vector of the block encoded using the inter-prediction is one.

However, if the motion vector of the current block is predicted more accurately, the motion vector can be encoded with a higher compression ratio. For this purpose, one embodiment of the present invention has a plurality of predicted motion vectors. By selecting one of the candidates and using it as the predicted motion vector of the current block, the motion vector of the current block is encoded with a higher compression ratio. In the following, a method of encoding the motion vector of the current block by using a plurality of predicted motion vector candidates will be described in more detail.

10A and 10B illustrate candidates for predicted motion vectors according to an embodiment of the present invention.

Referring to FIG. 10A, the method of predicting a motion vector according to an embodiment of the present invention uses one of the motion vectors of a previously encoded block adjacent to the current block as the predicted motion vector of the current block. can do. Of the blocks adjacent to the top of the current block, the leftmost a0 block, the top b0 block adjacent to the left side, the c block adjacent to the upper right side, the d block adjacent to the upper left side, and the e block adjacent to the lower left side. Any of the motion vectors of can be used as candidates for the predicted motion vector of the current block.

The video coding method and decoding method according to the present invention perform video coding and decoding based on coding units of various sizes classified by depth, and the movement of the e-block adjacent to the lower left side. Vectors can also be used as predictive motion vector candidates.

Explaining with reference to FIGS. 8A and 8B, if the current block is the coding unit 820, then the top, left top, right top, left and bottom left coding units 814,816,818 of the current block and 822 is currently encoded before the block. Therefore, the motion vector of the block adjacent to the lower left side of the current block can also be used as a predicted motion vector candidate.

With reference to FIG. 10B, the motion vectors of all the blocks currently in contact with each other can be used as candidates for the predicted motion vector. In other words, not only the leftmost a0 block among the blocks adjacent to the upper part but also the motion vectors of all the blocks (a0 to aN) adjacent to the upper part can be used as predictive motion vector candidates, and are adjacent to the left side. Among the blocks, not only the uppermost b0 block but also the motion vectors of all the blocks (b0 to bN) adjacent to the left side can be used as the predicted motion vector candidates.

In addition, the median motion vector of adjacent blocks can be used as a predicted motion vector candidate. In other words, median (mv_a0, mv_b0, mv_c) can be used as a predicted motion vector candidate for the current block. Here, mv_a0 is a motion vector of a0 block, mv_b0 is a motion vector of b0 block, and mv_c is a motion vector of c block.

However, the candidates for the predicted motion vector of the current block can be limited by the size of the current block and the size of the adjacent block, which will be described in detail with reference to FIGS. 10C to 10E.

10C to 10E illustrate blocks of various sizes adjacent to the current block according to one embodiment of the invention.

As described above, the video coding and decoding methods according to the present invention encode video using coding units and prediction units of various sizes determined by depth. Therefore, the sizes of the blocks adjacent to the current block are also various, but if the size of the current block and the size of the partially adjacent block are significantly different, the motion vector of the partially adjacent blocks with different sizes May not be used as a predicted motion vector candidate.

Referring to FIG. 10C, the blocks 1014 to 1018 adjacent to the upper part of the current block 1010 are blocks smaller than the size of the current block 1010. Since it is highly possible that the motion vector of the adjacent block 1012, which is currently the same size as the block 1010, is the same as or similar to the motion vector of the current block 1010, the motion vector estimation unit 910 has the same size. Only the motion vectors of adjacent blocks 1012 can be used as predictive motion vector candidates.

Even if the sizes are not the same, only motion vectors of adjacent blocks of a predetermined size or larger can be used as predicted motion vector candidates. For example, only motion vectors of blocks (1012 and 1018) having a size of 1/4 or more of the size of the current block 1010 can be used as predicted motion vector candidates.

Referring to FIG. 10D, the size of the block 1022 adjacent to the left side of the current block 1020 is 16 times that of the current block, and there is a significant difference in size. Due to the significant difference in magnitude, it is unlikely that the motion vector of block 1022 adjacent to the left side is currently the same as or similar to the motion vector of block 1020. Therefore, the motion vector of the block 1022 adjacent to the left side is not currently used as the predicted motion vector candidate of the block 1020, and only the motion vector of the block 1024 adjacent to the upper part and the block 1026 adjacent to the upper left side is used as the predicted motion vector candidate. can do.

Referring to FIG. 10E, the current block (the size of 1030 is larger than the size of all adjacent blocks 1031 to 1037. At this time, the motion vectors of all adjacent blocks 1031 to 1037 are all the motion vectors of the current block 1030. If it is used as a predicted motion vector candidate, the number of predicted motion vector candidates in the current block 1030 is too large. The larger the size difference between the current block 1030 and the adjacent blocks 1031 to 1037, the more the candidate for the predicted motion vector. Therefore, the motion vector estimation unit 910 according to the embodiment of the present invention does not currently use the motion vector of a part of the adjacent blocks as the predicted motion vector candidate of the block 1030.

For example, in the embodiment illustrated in FIG. 10E, the motion vectors of block 1031 adjacent to the lower left side and block 1037 adjacent to the upper right side may not be currently used as predicted motion vector candidates for block 1030.

This is further generalized, and if the size of the current block 1030 is equal to or larger than a predetermined size, the motion vector of the adjacent block in a specific direction is not used as the predicted motion vector candidate of the current block 1030. Sometimes.

11A to 11C illustrate candidates for predicted motion vectors according to other embodiments of the present invention.

FIG. 11A illustrates a method of determining a predictive motion vector candidate for a B-picture (bi-directional predictive picture) according to an embodiment of the present invention. When the current picture including the current block is a B picture that performs bidirectional prediction, the motion vector generated based on the temporal distance may be a predicted motion vector candidate.

The predicted motion vector candidate (mv_temporal) of the current block 1100 of the current picture 1110 is determined by using the motion vector of the block 1120 at the same position (colocated) as the time-preceding picture 1112. For example, if the motion vector mv_colA of the block 1120 at the same position as the current block 1100 is generated for the searched block 1122 of the picture 1114 that currently follows the time of the picture 1110, the predicted motion vector of the current block 1100. Mv_L0A and mv_L1A, which are candidates for, are determined as follows.

mv_L1A = (t1 / t2) * mv_colA
mv_L0A = mv_L1A-mv_colA
Here, mv_L0A means a predicted motion vector candidate of the current block 1110 related to the time-preceding picture 1112, and mv_L1A means a predicted motion vector candidate of the current block 1110 related to the time-following picture 1114. do.

In the embodiment illustrated in FIG. 11A, the current picture 1110, which is a B picture, exists between the time-leading picture 1112 and the time-following picture 1114. At this time, if the motion vector mv_colA of the block 1120 at the same position is generated for the time-following picture 1114 of the current picture 1110, the motion vector of the current block 1100 is predicted more accurately based on mv_L1A. can do. In other words, mv_colA is more than when mv_colA is a motion vector in the direction opposite to the direction shown in FIG. 11A, that is, when it is generated for other pictures prior to picture 1112 which precedes in time. With the motion vector in the direction illustrated in 11A, the motion vector of block 1100 can now be predicted more accurately.

Therefore, if the direction from the current block 1110 to the block 1120 at the same position is the List0 direction, the motion vector mv_colA of the block 1120 at the same position must be in the List1 direction, as shown in FIG. 11A. The current picture 1110 is more likely to exist between the preceding picture 1112 and the trailing picture 1114, and the motion vector of the current block 1100 can be more accurately predicted based on mv_colA.

Further, since the pictures 1110 to 1114 illustrated in FIG. 11A are arranged over time, it is possible to generate a predicted motion vector candidate (mv_temporal) of the current block based on the POC (picture order count). Since the picture referenced by the current block may be other pictures other than the adjacent pictures 1112 and 1114 illustrated in FIG. 11A, a predicted motion vector candidate for the current block is generated based on the POC.

For example, if the POC of the current picture is CurrPOC and the POC of the picture currently referenced by the picture is CurrRefPOC, the predicted motion vector candidates for the current block are also generated as follows.

Scale = (CurrPOC-CurRefPOC) / (ColPOC-ColRefPOC)
mv_temporal = Scale * mv_colA
Here, the ColPOC is the POC of the picture 1112 that precedes the block 1120 at the same position in time, and the ColRefPOC is the POC of the block 1122 that is referenced by the block 1120 at the same position later in time. The POC of picture 1114 to be performed.

FIG. 11B illustrates a method of generating a predicted motion vector candidate for a B picture according to another embodiment of the present invention. Compared with the method illustrated in FIG. 11A, picture 1114, which follows in time, is different in that a block at the same position as block 1100 currently exists.

Referring to FIG. 11B, the predicted motion vector candidate of the current block 1100 of the current picture 1110 is also generated by using the motion vector of the block 1130 at the same position as the time-following picture 1114. For example, if the motion vector mv_colB of the block 1130 at the same position as the current block 1100 is generated for the searched block 1132 of the picture 1112 that precedes the current picture 1110 in time, the predicted motion vector of the current block 1100 will be generated. Candidates mv_L0B and mv_L1B are also generated as follows.

mv_L0B = (t3 / t4) * mv_colB
mv_L1B = mv_L0B-mv_colB
Here, mv_L0B means a predicted motion vector of the current block 1110 related to the time-preceding picture 1112, and mv_L1B means a predicted motion vector candidate of the current block 1100 related to the time-following picture 1114. ..

Similar to FIG. 11A, also in the embodiment illustrated in FIG. 11B, the current picture 1110, which is a B picture, exists between the time-leading picture 1112 and the time-following picture 1114. Therefore, if the motion vector mv_colB of the block 1130 at the same position is generated for the picture 1112 that precedes in time, the motion vector of the current block 1100 can be predicted more accurately based on the mv_L0B. In other words, mv_colB is more than if mv_colB is a motion vector in the direction opposite to the direction shown in FIG. 11B, that is, it is generated for another picture after picture 1114 that follows in time. With the motion vector in the direction illustrated in 11B, the motion vector of block 1100 can now be predicted more accurately.

Therefore, if the direction from the current block 1110 to the block 1130 at the same position is the List1 direction, the motion vector mv_colB of the block 1130 at the same position must be in the List0 direction, as shown in FIG. 11B. The current picture 1110 is more likely to exist between the preceding picture 1112 and the trailing picture 1114, and the motion vector of the current block 1100 can be predicted more accurately based on mv_colB.

Further, since the picture referred to by the current block may be another picture other than the adjacent pictures 1112 and 1114 shown in FIG. 11B, the predicted motion vector candidate of the current block is generated based on the POC.

For example, if the POC of the current picture is CurrPOC and the POC of the picture currently referenced by the picture is CurrRefPOC, the predicted motion vector candidates for the current block are also generated as follows.

Scale = (CurrPOC-CurRefPOC) / (ColPOC-ColRefPOC)
mv_temporal = Scale * mv_colB
Here, the ColPOC is the POC of the picture 1114 that follows the time when the block 1130 at the same position is included, and the ColRefPOC is the POC of the block 1132 that the block 1130 at the same position refers to. The POC of the preceding picture 1112.

FIG. 11C illustrates a predictive motion vector candidate for a P picture (predictive picture) according to an embodiment of the present invention. With reference to FIG. 11C, the predicted motion vector candidate of the current block 1100 of the current picture 1110 is determined by using the motion vector of the block 1140 at the same position as the time-preceding picture 1112. For example, if the motion vector mv_colC of the block 1140 at the same position as the current block 1100 is generated for the searched block 1142 of the other time-preceding picture 1116, it is a predicted motion vector candidate of the current block 1100. mv_L0C is determined as follows.

mv_L0C = (t6 / t5) * mv_colC
As described in connection with FIGS. 11A and 11B, mv_L0C can also be determined based on POC. The mv_L0C can be determined based on the POC of the current picture 1110, the POC of the picture currently referenced by the picture 1110, the POC of the time-preceding picture 1112, and the POC of the other time-preceding picture 1116.

Since the current picture 1110 is a P picture, only one predicted motion vector candidate of the current block 1100 is determined unlike FIGS. 11A and 11B.

Further, in FIGS. 11A and 11B, in order to use the predicted motion vector candidate generated based on the temporal distance for predicting the motion vector of the current block, any of the blocks 1120 and 1130 at the same position can be used. Information indicating whether to generate a predicted motion vector candidate using the block of the above must be encoded together, and this information is included in the slice header or the sequence header as a slice parameter or a sequence parameter.

In summary, according to FIGS. 10A and 10B, FIGS. 11A to 11C, the set C of candidates for the predicted motion vector is as follows.

C = {median (mv_a0, mv_b0, mv_c), mv_a0, mv_a1, ..., mv_aN, mv_b0, mv_b1, ..., mv_bN, mv_c, mv_d, mv_e, mv_temporal}
Alternatively, the set C may be a set in which the number of candidates for the predicted motion vector is reduced.

C = {median (mv_a', mv_b', mv_c'), mv_a', mv_b', mv_c', mv_temporal}
Here, mv_x means the motion vector of the x block, median () means the median value, and mv_temporal is the prediction generated by using the temporal distance described in connection with FIGS. 11A to 11C. It means a motion vector candidate. mv_a'means the first valid motion vector of mv_a0, mv_a1, ..., Mv_aN. For example, if the a0 block is encoded using intra-prediction or refers to a picture different from the current block, the motion vector of a0, mv_a0, is not valid, so mv_a'= mv_a1. , When the motion vector of the a1 block is also not valid, mv_a'= mv_a2. Similarly, mv_b'means the first valid motion vector of mv_b0, mv_b1, ..., Mv_bN, and mv_c'means the first valid motion vector of mv_c, mv_d, mv_e.

Of the motion vectors of the blocks adjacent to the current block, the motion vector of the block that refers to a picture different from the current block cannot efficiently predict the motion vector of the current block. Therefore, in the set of predicted motion vector candidates C, the motion vector of a block that refers to a picture different from the current block can be excluded.

When the motion vector coding apparatus 900 encodes the motion vector in the explicit mode, it indicates which of the C sets, the predicted motion vector candidates, is used to predict the motion vector of the current block (signaling). ) Information is also encoded. In other words, when the motion vector encoder 900 encodes a motion vector, it assigns a binary number corresponding to each element of the C set, that is, a candidate for the predicted motion vector, and uses it to predict the motion vector of the current block. The binary numbers corresponding to the predicted motion vector candidates are also encoded.

In order to identify one of the elements of the C set, a binary number corresponding to each predicted motion vector candidate is assigned and the binary number is output. Therefore, the smaller the number of elements in the C set, the smaller the number of bits. The element of the C set can be specified by the binary number.

Therefore, if there are overlapping predicted motion vector candidates in the C set, the duplicated predicted motion vector candidates are excluded from the C set and a binary number can be assigned. For example, when the C set is C = {median (mv_a', mv_b', mv_c'), mv_a', mv_b', mv_c', mv_temporal} as described above, mv_a', mv_b' and mv_c' are If they are all the same, the C set can be determined by two elements, such as C = {mv_a', mv_temporal}, and a binary number can be assigned. If the five elements of the C set can be identified using 3 bits before excluding duplicate predicted motion vector candidates, then two after excluding duplicate predicted motion vector candidates. The element of can be specified by using 1 bit.

Instead of excluding duplicate predicted motion vector candidates, a predetermined weight may be added to increase the probability that the duplicate predicted motion vector candidates will be determined as the predicted motion vector of the current block. can. In the above example, mv_a', mv_b', and mv_c'are all the same, and only mv_a'is included in the C set. Therefore, a predetermined weighted value is added to mv_a', and mv_a'is the motion vector of the current block. Can be used to increase the probability of being used to predict.

Further, when there is only one predicted motion vector candidate, the binary number for specifying one of the predicted motion vector candidates may not be encoded. For example, the C set is C = {median (mv_a0, mv_b0, mv_c), mv_a0, mv_a1, ..., mv_aN, mv_b0, mv_b1, ..., mv_bN, mv_c, mv_d, mv_e, mv_temporal}. If the bN block, c block, d block, and e block are all intra-predicted blocks, the C set contains C = {mv_temporal}, and therefore contains substantially only one element. Therefore, in this case, the motion vector coding device 900 may not encode a binary number for identifying one of the predicted motion vector candidates.

It can be easily understood by those skilled in the art to which the present invention belongs that other motion vectors other than all the candidates for the predicted motion vector described above are used as candidates for the predicted motion vector. Will.

Further, according to another embodiment of the present invention, the candidate determination unit 920 can reduce the number of candidates for the predicted motion vector.

As mentioned above, of the multiple predictive motion vector candidates, separate information is encoded and included in the bitstream to identify the predictive motion vector candidates used to predict the motion vector of the current block. Is done. Therefore, the smaller the number of elements in the C set, the less bits the information needed to identify the predicted motion vector candidates currently used to predict the motion vector of the block in the C set. do. Therefore, the candidate determination unit 920 can selectively exclude a predetermined predicted motion vector candidate from the total candidates of the predicted motion vector by using a predetermined evaluation function. This will be described in detail with reference to FIG.

FIG. 12 illustrates a method of reducing the number of predicted motion vector candidates according to an embodiment of the present invention. In FIG. 12, it is assumed that the number of elements in the C set is three, MVP1, MVP2 and MVP3 are the elements in the C set, and the motion vector of the block is currently MV. Since the predicted motion vector candidates most similar to the motion vector of the current block are used to predict the motion vector of the current block, MVP3 most similar to the MV is used to predict the motion vector of the current block. NS.

Therefore, in the motion vector coding apparatus 900, the difference vector between the motion vector of the current block encoded as information about the motion vector and the predicted motion vector candidate used for predicting the motion vector of the current block (hereinafter referred to as , "Actual motion vector difference") is (2,0). Since the MV is (5,0) and the MVP3 is (3,0), the actual motion vector difference is (2,0).

The candidate determination unit 920 selectively excludes at least one predicted motion vector candidate from the total candidates of the predicted motion vector by utilizing such an actual motion vector difference and a predetermined evaluation function. More specifically, a virtual motion vector is generated by using the actual motion vector difference and a predetermined predicted motion vector candidate, and the motion vector difference between the generated virtual motion vector and the overall candidate (hereinafter referred to as “)”. "Virtual motion vector difference") is generated for all candidates. A virtual motion vector is generated by adding the actual motion vector difference and a predetermined predicted motion vector candidate, and the motion vector difference between the generated virtual motion vector and the overall candidate is calculated. By comparing the actual motion vector difference with the virtual motion vector difference calculated for each of the overall candidates, a predetermined predicted motion vector candidate can be selectively excluded from the overall candidates of the predicted motion vector.

To explain in detail with reference to FIG. 12, first, the candidate determination unit 920 determines whether or not to exclude MVP1, which is one of the predicted motion vector candidates, from the overall candidates.

If the virtual motion vector difference generated by subtracting the predicted motion vector candidates different from the virtual motion vector based on MVP1 is smaller than the actual motion vector difference, MVP1 is used to predict the motion vector of the current block. Not done. For example, if the virtual motion vector difference generated by subtracting MVP3 from the virtual motion vector generated by adding the MVP1 and the actual motion vector difference is smaller than the actual motion vector difference, the MVP3 is larger than the MVP1. By predicting the virtual motion vector more accurately, it is clear that the MVP1 cannot be the predicted motion vector in this case.

In FIG. 12, if the MVP1 and the actual motion vector difference are added, the virtual motion vector based on the MVP1 is (2,0). Therefore, when a virtual motion vector is generated based on MVP1, the virtual motion vector difference related to MVP2 is (2,0), and the virtual motion vector difference related to MVP3 is (-1,0). Is. At this time, since the magnitude of the virtual motion vector difference (-1,0) related to the MVP3 is smaller than the magnitude of the actual motion vector difference (2,0), the MVP1 is the predicted motion of the current block. Can't be a vector. Therefore, MVP1 can be excluded from the overall candidates for the predicted motion vector. In other words, in the above-mentioned C set, the predicted motion vector candidates corresponding to MVP1 may be excluded.

At this time, the virtual motion vector difference calculated for the MVP1 itself is (2,0), which is always the same as the actual motion vector difference, and therefore cannot be smaller than the magnitude of the actual motion vector. Therefore, when calculating the virtual motion vector difference for each of the overall candidates of the predicted motion vector, the virtual motion vector difference related to the MVP1 itself is not calculated.

When the determination of exclusion of MVP1 is completed, the candidate determination unit 920 determines whether or not to exclude MVP2 from all the candidates of the predicted motion vector. If MVP2 is added to the actual motion vector difference, the virtual motion vector based on MVP2 is (2,0). Therefore, the virtual motion vector difference related to MVP1 is (2,0), and the virtual motion vector difference related to MVP3 is (-1,0). Since the magnitude of the virtual motion vector difference related to MVP3 is smaller than the magnitude of the actual motion vector difference, MVP2 is excluded from the overall candidates of the predicted motion vector as well as MVP1. When determining the exclusion of MVP2, the comparison between the virtual motion vector difference related to MVP1 and the actual motion vector difference is selective. Since it was determined that MVP1 is no longer the predicted motion vector of the current block, it is possible to compare the virtual motion vector difference for each of the remaining candidates excluding MVP1 with the actual motion vector difference. can.

The candidate determination unit 920 also determines whether to exclude MVP3. The virtual motion vector based on MVP3 is the same as the actual motion vector, and even if other predicted motion vector candidates (that is, MVP1 or MVP2) are subtracted from the actual motion vector, it is smaller than the magnitude of the actual motion vector difference. Since no virtual motion vector difference occurs, MVP3 is not excluded from the overall candidates for the predicted motion vector. Also, according to other embodiments of the present invention, it is clear that MVP3, which is currently determined to be used to predict the motion vector of the block, is not excluded from the overall candidates for the predicted motion vector. The determination unit 920 can skip exclusions for the predicted motion vector candidates that are currently determined to be used to predict the motion vector of the block.

In short, the candidate determination unit 920 determines whether to exclude the second predicted motion vector, which is one of the overall candidates for the predicted motion vector, and adds the second predicted motion vector and the difference between the actual motion vectors. A virtual motion vector is generated, a difference vector of a predicted motion vector different from the virtual motion vector is calculated for each of all candidates, and a plurality of virtual motion vector differences are generated. If there is at least one virtual motion vector difference smaller than the actual motion vector difference among the plurality of virtual motion vector differences, the second predicted motion vector may not be the predicted motion vector of the current block. Since it is obvious, it is excluded from the overall candidates for the predicted motion vector.

Further, the candidate determination unit 920 reduces the total number of candidates for the predicted motion vector, that is, the number of elements in the C set, by repeating the judgment on such exclusions for each of the overall candidates for the predicted motion vector. be able to. The exclusion is determined in order according to the well-ordered order of the candidates of the overall predicted motion vector of the C set. For example, when C = {median (mv_a', mv_b', mv_c'), mv_a', mv_b', mv_c', mv_temporal}, the exclusion of median (mv_a', mv_b', mv_') is determined. When the judgment is completed, the exclusion of mv_a'is judged. After that, the exclusion of mv_b'is determined. Depending on the order of the C set, the exclusion judgment is repeated up to mv_temporal.

It is explained in relation to the judgment of MVP2 exclusion that it is possible to omit the comparison between the virtual motion vector difference and the actual motion vector difference for the candidates excluded in the previous judgment process when making an iterative judgment. That's right.

Further, the C set is rearranged according to a predetermined criterion as described later in relation to FIGS. 13A to 13D, but when the C set is rearranged, the exclusion is performed according to the rearranged order. Repeat the judgment of.

The magnitude comparison between the virtual motion vector difference and the actual motion vector difference described with reference to FIG. 12 is applied not only to the one-dimensional motion vector but also to the two-dimensional motion vector. In other words, the magnitude of the virtual motion vector difference defined by the x-coordinate and the y-coordinate can be compared with the magnitude of the actual motion vector difference, and a predetermined predicted motion vector candidate can be selectively excluded from the overall candidates. can.

However, the magnitude that is the criterion for comparison between the virtual motion vector difference and the actual motion vector difference is an example, and various criteria are the virtual motion vector difference and the actual motion vector difference. It is also used for comparison with. When the evaluation function that generates the value related to the virtual motion vector difference and the value related to the actual motion vector difference is "A" based on a predetermined standard, the virtual motion vector difference is calculated by the following mathematical formula (1). And the actual motion vector difference can be compared.

A (mvx + MVD-mvy) <A (MVD) (1)
The candidate determination unit 920 satisfies the mathematical formula (1) in order to determine whether or not to exclude "mvx", which is one of the overall candidates of the predicted motion vector, from the overall candidates of the predicted motion vector. Determine if at least one is present in the overall candidates. In the formula (1), "MVD" means the actual motion vector difference. In order to determine the exclusion of "mvx", the virtual motion vector difference between "mvx + MVD", which is a virtual motion vector based on "mvx", and "mvy", which is a predicted motion vector candidate, is "mvx + MVD-". "Mvy" is calculated by calculating "A (mvx + MVD-mvy)" which is a value evaluated by using a predetermined evaluation function "A", and the value generated as a result of the calculation is a value related to the actual motion vector difference. Compare with a certain "A (MVD)". Among the overall candidates, there is at least one "mvy" that satisfies the formula (1) by repeatedly substituting the other predicted motion vector candidates excluding "mvx" into "mvy". Judge whether or not.

As described above, the virtual motion vector difference and the actual motion vector difference evaluated by "A" are also defined by the x-coordinate and the y-coordinate. In this case, the evaluation function is also defined as the sum of the evaluated values of the x-coordinate and the evaluated value of the y-coordinate, as in the following mathematical formula (2).

A (p, q) = f (p) + f (q) (2)
When the virtual motion vector difference or the actual motion vector difference is defined as the x-coordinate "p" and the y-coordinate "q", the respective coordinate values are assigned to the predetermined function "f", and the sum of the assigned results is used. , The evaluation function "A" is also defined.

According to one embodiment of the present invention, the evaluation function "A" of the mathematical formulas (1) and (2) is the result of entropy-encoding the virtual motion vector difference and the result of entropy-coding the actual motion vector difference. It is an evaluation function that estimates. The candidate determination unit 920 estimates the result of entropy-encoding the virtual motion vector difference and the actual motion vector difference based on the evaluation function "A", and reduces the number of predicted motion vector candidates based on the estimation result. be able to. This will be described in detail with reference to the mathematical formula (3).

Length = 1;
Temp = (val << = 0 (-val << 1) + 1: (val <<1);
While (1! = Temp) {
Temp >> = 1;
Length + = 2;
}
f (val) = Length (3)
The function "f" for estimating the entropy-encoding result for the x-coordinate value or the y-coordinate value is also defined as in the equation (3). If the function "f" that predicts the result of variable length coding (for example, universal variable length coding) is entered with "val", which is an x-coordinate value or a y-coordinate value. , "Length" is calculated by the above formula (3).

Formula (3) can also be expressed as follows.

ｘ座標値またはｙ座標価格は、仮想の動きベクトル差または実際動きベクトル差のｘ座標値またはｙ座標値であってもよい。

The x-coordinate value or the y-coordinate price may be the x-coordinate value or the y-coordinate value of the virtual motion vector difference or the actual motion vector difference.

According to the formula (3), if "val" is a negative number or "0", after converting "val" to a positive number, shift it to the left by about 1 bit, multiply the coordinate value by "2", and then multiply. Add "1" and save as "Temp". If "val" is a positive number, shift "val" to the left by about 1 bit, multiply the coordinate value by "2", and save it as "Temp". After that, the "while" loop is repeated until "Temp" becomes "1", and "Length" is calculated.

For example, if the virtual motion vector difference or the actual motion vector difference is (2,0), then A (2,0) = f (2 + f (0).

f (2) is calculated as follows. Since "2" in f (2) is a positive number, shift it to the left by about 1 bit and set "Temp" to "4". In the first while loop, "Temp" is "4" and "4" is not "1", so by shifting "4" to the right and multiplying by "1/2", "Temp" Is set to "2". Since the initial value of "Length" is set to "1", "Length" becomes "3" in the first while loop.

In the second while loop, "Temp" is "2" and "2" is not "1", so by shifting "2" to the right and multiplying by "1/2", "Temp" Is set to "1". Currently, "Length" is "3", so in the second while loop, "Length" becomes "5". The third while loop is not executed because "Temp" is "1", and f (2) becomes "5".

f (0) is calculated as follows. Since the input coordinate value of f (0) is "0", "0" is shifted to the left by 1 bit, "1" is added, and "Temp" is set to "1". Therefore, the while loop is not executed. Depending on the initial value of "Length", f (0) becomes "1".

The predetermined evaluation function "f" described in relation to the mathematical formula (3) is a function for estimating the result of entropy coding using variable length coding. Therefore, the candidate determination unit 920 uses the evaluation function "A" to determine whether or not to exclude "mvx" from the overall candidates of the predicted motion vector, and makes the virtual motion vector difference variable length forever. Estimate the encoded result. If there is at least one virtual motion vector difference estimated to be encoded to a length shorter than the actual motion vector difference as a result of the estimation, "mvx" is excluded from the overall candidates for the predicted motion vector.

However, those skilled in the art to which the present invention belongs will readily appreciate that entropy coding results by other methods that are not the result of variable length coding can be estimated. For example, another evaluation function "h" can be used to estimate and compare the entropy coding result of the virtual motion vector difference and the entropy coding result of the actual motion vector difference. "h" may be a function that estimates the result of context adaptive binary arithmetic coding.

Further, according to another embodiment of the present invention, the result of evaluating the index information can be estimated together in order to improve the accuracy of the evaluation result based on the predetermined evaluation function. The index information is information for identifying a predetermined predicted motion vector candidate among the overall candidates of the predicted motion vector. This will be described in detail with reference to the mathematical formula (4).

A (mvx + MVD-mvy, mbyIdx) <A (MVD, mbxIdx) (4)
The candidate determination unit 920 satisfies the mathematical formula (4) in order to determine whether or not to exclude "mvx", which is one of the overall candidates of the predicted motion vector, from the overall candidates of the predicted motion vector, "mvy". However, it is determined whether or not at least one of the overall candidates exists. In the formula (4), "MVD" means the actual motion vector difference, and mbxIdx and mbyIdx are all candidates of the predicted motion vector, and mean the index information for specifying "mvx" and "mvy", respectively. .. In order to determine the exclusion of "mvx", the virtual motion vector difference between "mvx + MVD", which is a virtual motion vector based on "mvx", and "mvy", which is a predicted motion vector candidate, is different from "mvx + MVD-". Index information for specifying "mvy" and "mvy" as a whole candidate is evaluated using a predetermined evaluation function "A", and actual motion vector difference and index information for specifying "mvx" as a whole candidate are evaluated. Is evaluated using a predetermined evaluation function "A". As a result of the evaluation, it is determined whether or not at least one "mvy" satisfying the mathematical formula (4) exists in the overall candidates.

As described above, the virtual motion vector difference and the actual motion vector difference evaluated by "A" are defined by the x-coordinate and the y-coordinate, and are also defined by the mathematical formula (5).

A (mvx + MVD-mvy, mvyIdx) = f (p1) + f (q1) + g (mvyIdx)
A (MVD, mbxIdx) = f (p2 + f (q2 + g (mvxIdx)) (5)
Compared with the formula (2), A (mvx + MVD-mvy) on the left side of the formula (2) evaluated only the virtual motion vector difference, but A (mvx + MVD-mvy, mvyIdx) of the formula (5) was virtual. The information for specifying "mvy" in the motion vector difference of and the overall candidate of the predicted motion vector is also evaluated. As described above, the evaluation function "A" may be a function for evaluating the result of entropy coding, and at this time, the function "f" is described in relation to the mathematical formula (2). It is a function for estimating the entropy coding result based on the x-coordinate value or the y-coordinate value of the virtual motion vector difference, and the function "g" is a function for estimating the entropy coding result of "mvxIdx". It may be. When the x-coordinate value of "mvx + MVD-mvy" is "p1" and the y-coordinate value is "q1", A (mvx + MVD-mvy, mbxIdx) shall be calculated as shown in the mathematical formula (5). Can be done.

A (MVD) on the right side of the equation (2) was also evaluated only for the actual motion vector difference, but A (MVD, mVxIdx) in the equation (5) is the overall candidate for the actual motion vector difference and the predicted motion vector, and is "mVx. Evaluate together the information to identify. The function "f" is a function for estimating the entropy coding result based on the x-coordinate value or the y-coordinate value of the actual motion vector difference, as explained in relation to the mathematical formula (2), and the function "g". May be a function for estimating the entropy coding result of "mvxIdx". When the x-coordinate value of "MVD" is "p2" and the y-coordinate value is "q2", A (MVD, mbxIdx) can be calculated as shown in the mathematical formula (5).

The exclusion judgment by the mathematical formulas (4) and (5) is also used as an auxiliary to the exclusion judgment by the mathematical formula (2). In other words, based on the mathematical formula (2), it is first determined whether or not to exclude "mvx" from the candidates for the overall predicted motion vector, and then the mathematical formulas (4) and (5) are used to supplement the exclusion once more. You can judge. For example, as a result of judgment by the mathematical formula (2), "A (mvx + MVD-mvy)" exists only when it is the same as or larger than "A (MVD)", and "A (mvx + MVD-mvy)" is present. If there is no case smaller than "A (MVD)", then according to formula (2), "mvx" is not excluded from the overall candidates for the predicted motion vector. However, even if "A (mvx + MVD-mvy)" and "A (MVD)" are equal, "mvx" is excluded from the overall candidates for the predicted motion vector based on the judgment results by the mathematical formulas (4) and (5). can do.

When the candidate determination unit 920 determines whether to exclude the predicted motion vector candidates based on the mathematical formulas (1) to (5), iterates over all the predicted motion vector candidates according to the well-ordered order of the C set. Was explained. According to another embodiment of the present invention, the candidate determination unit 920 can rearrange the C set according to a predetermined criterion, and can repeat the determination of exclusion depending on the rearrangement order. This will be described in detail with reference to FIGS. 13A to 13D.

13A to 13D illustrate the position of the current block contained in a coding unit of a predetermined size according to one embodiment of the present invention.

When the candidates for the overall predicted motion vector are C = {median (mv_a', mv_b', mv_c'), mv_a', mv_b', mv_c', mv_temporal}, each of the predicted motion vector candidates in the C set is a binary number. It was explained that there is a possibility of identifying the predicted motion vector used to predict the motion vector of the current block among the candidates of the predicted motion vector by assigning.

At this time, a binary number is assigned according to the alignment order of the candidates of the predicted motion vector included in the C set, and such a binary number may be a variable length code based on the Huffman code. Therefore, a smaller number of bits can be assigned to the previously located predicted motion vector candidates in the well-ordered C set. For example, in the C set, the "median (mv_a', mv_b', mv_c')" can be assigned the "0" bit, the mv_a'can be assigned the "00" bit, and the mv_b'can be assigned the "01" bit. Therefore, the candidate determination unit 920 predicts that among the candidates for the predicted motion vector, the predicted motion vector candidate that is likely to be used for predicting the motion vector of the current block is located in front of the C set. Motion vector candidates are aligned in a predetermined order.

The predicted motion vector, which is likely to be used to predict the motion vector of the current block, is the coding unit and is determined by the position of the current block. If the current block is located at the lower end of the coding unit, as in FIG. 13A, the motion vector of the current block will be the motion vector of the block adjacent to the left side of the coding unit or the motion vector of the block adjacent to the lower left side. Most likely the same or similar. Therefore, it is necessary to change the well-order so that the motion vector of the block adjacent to the left side or the candidate of the predicted motion vector corresponding to the motion vector of the block adjacent to the lower left side is located in front of the C set. Since mv_b'among the candidates for the predicted motion vector of the C set described above is a candidate for the predicted motion vector corresponding to the motion vector of the block adjacent to the left side, the C set is mv_b'and median (mv_a', mv_b'. , Mv_c'), and rearranged as C = {mv_b', mv_a', median (mv_a', mv_b', mv_c'), mv_c', mv_temporal}.

Similarly, as shown in FIG. 13B, if the current block is located on the left side of the coding unit, the motion vector of the block adjacent to the left side of the coding unit and the predicted movement corresponding to the motion vector of the block adjacent to the upper part. Vector candidates are likely to be used to predict the motion vector of the current block. Of the above-mentioned candidates for the predicted motion vector of the C set, mv_b'is a predicted motion vector candidate corresponding to the motion vector of the block adjacent to the left side, so that the C set is mv_b'and median (mv_a', mv_b', The order with mv_c') is changed, and the order is also rearranged as C = {mv_b', mv_a', median (mv_a', mv_b', mv_c'), mv_c', mv_temporal}.

As shown in FIG. 13C, if the current block is located above the coding unit, the motion vector of the block adjacent to the left side of the coding unit and the predicted motion vector candidate corresponding to the motion vector of the block adjacent to the top can be obtained. Currently, it is likely to be used for the predicted motion vector of the block. Of the above-mentioned candidates for the predicted motion vector of the C set, mv_a'is the predicted motion vector candidate corresponding to the motion vector of the block adjacent to the upper part, so that the C set is mv_a'and median (mv_a', mv_b', The order with mv_c') is changed, and the order is also rearranged as C = {mv_a', median (mv_a', mv_b', mv_c'), mv_b', mv_c', mv_temporal}.

As shown in FIG. 13D, if the current block is located on the right side of the coding unit, the predicted motion vector candidate corresponding to the motion vector of the block adjacent to the upper right side of the coding unit predicts the motion vector of the current block. It is likely to be used for. Of the above-mentioned candidates for the predicted motion vector of the C set, mv_c'is a predicted motion vector candidate corresponding to the motion vector of the block adjacent to the upper right side, so that the C set is mv_c'and median (mv_a', mv_b'. , Mv_c'), and rearranged as C = {mv_c', mv_a', mv_b', median (mv_a', mv_b', mv_c'), mv_temporal}.

The position of the current block in the coding unit is exemplary as a criterion for rearranging the candidates for the predicted motion vector. In other words, various criteria are also used as criteria for rearranging the candidates for the predicted motion vector. Various criteria for aligning predicted motion vector candidates, which are likely to be similar to the motion vector of the current block, to the front of the C set are also used as criteria for rearranging the predicted motion vector candidates. Based on certain information related to other blocks encoded before the current block, predictive motion vector candidates that are likely to be similar to the motion vector of the current block are determined, and the C set is rearranged based on the determination. can do.

Also, a predicted motion vector candidate that is likely to be similar to the motion vector of the current block, based on other information encoded or decoded for the current block before encoding the motion vector of the current block. Can be determined and the C set can be rearranged based on the determination.

In addition, the predicted motion vector candidates that are duplicated when the C set is rearranged are excluded, and the rearrangement can be performed. If there are duplicate predicted motion vector candidates in the overall candidates for the predicted motion vector, they are first excluded, and each predicted motion vector candidate is excluded by the above formulas (1) to (5). I can judge the situation.

Referring to FIG. 9 again, the motion vector coding unit 930 encodes the information about the motion vector and the information about the predicted motion vector. The information about the motion vector is the difference vector between the actual motion vector of the current block and the actual predicted motion vector, and the information about the predicted motion vector includes at least one predicted motion vector among the overall candidates of the predicted motion vector. The excluded candidates encode the information to identify the predicted motion vector used to predict the motion vector of the current block. In other words, the information for identifying the predicted motion vector of the current block in the candidate of the predicted motion vector that is not excluded from the candidate determination unit 920 is encoded as the information about the predicted motion vector.

The actual motion vector difference is received from the motion vector estimation unit 910, encoded by a predetermined entropy coding method, and the candidate determination unit 920 selectively excludes at least one predicted motion vector candidate to determine the predicted motion. The vector candidates encode the information for identifying the predicted motion vector candidates used to predict the motion vector of the current block determined by the motion vector estimation unit 910.

The candidate determination unit 920 determines the candidate for the predicted motion vector by excluding at least one candidate for the predicted motion vector from the overall candidates for the predicted motion vector by the above equations (1) to (5). Among the candidates for the predicted motion vector, the information for identifying the predicted motion vector candidate used for predicting the motion vector of the current block is encoded. The motion vector coding unit 930 can index each candidate of the predicted motion vector that is not excluded from the candidate determination unit 920, and can entropy-encode the index information as the information about the predicted motion vector. Indexing means assigning a predetermined binary number to each candidate of the predicted motion vector, and the information about the predicted motion vector is used to predict the motion vector of the current block among the candidates of the predicted motion vector. Means the information given. As a result of the candidate determination unit 920 selectively excluding at least one predicted motion vector candidate, if only one predicted motion vector candidate remains, the motion vector coding unit 930 separately encodes the information about the predicted motion vector. There is no need to change. This is because the predicted motion vector candidates used to predict the motion vector of the current block are implicitly determined.

Further, as described with reference to FIGS. 13A to 13D, the candidate determination unit 920 rearranges the candidates for the overall predicted motion vector according to a predetermined criterion, and at least one of the rearranged candidates for the overall predicted motion vector is selected. It is also possible to index each candidate of the predicted motion vector generated by selectively excluding one predicted motion vector and entropy-encode the index information.

As a result of the rearrangement of the candidate determination unit 920, the binary number with the smallest number of bits is assigned to the predicted motion vector candidate that is likely to be used to predict the motion vector of the current block. Information can be encoded with even higher compression ratios.

FIG. 14 illustrates an apparatus that decodes a motion vector according to an embodiment of the present invention.

A device included in the video decoding device 200 of FIG. 2 or the video coding unit 500 of FIG. 5 and decoding a motion vector is illustrated in detail in FIG. Referring to FIG. 14, the motion vector decoding apparatus 1400 according to the embodiment of the present invention includes a motion vector decoding unit 1410, a candidate determination unit 1420, and a motion vector restoration unit 1430.

The motion vector decoding device 1400 of FIG. 14 illustrates a device that decodes the motion vector of the current block when the motion vector of the current block is encoded by the explicit mode among the above-mentioned explicit modes and implicit modes. ing.

The motion vector decoding unit 1410 receives the bitstream related to the motion vector of the current block, and decodes the received bitstream. Decrypt the information about the motion vector contained in the bitstream. Decode the actual motion vector difference of the current block. The actual motion vector difference can be decoded by a predetermined entropy decoding method. The actual motion vector difference is the difference vector between the motion vector of the current block and the predicted motion vector candidate used to predict the motion vector of the current block.

According to the method of encoding the motion vector of the present invention, at least one predicted motion vector candidate is excluded from the total candidates of the predicted motion vector by the above-mentioned mathematical formulas (1) to (5), and the predicted motion vector is obtained. Candidates are decided. The candidates for the predicted motion vector are not fixed, but are continuously changed as the decoding proceeds in block units. Therefore, even if the information about the predicted motion vector candidates is the same, if the predicted motion vector candidates are not determined, the predicted motion vector candidates used to predict the motion vector of the current block are accurately restored. Can not do it.

Therefore, prior to determining the predicted motion vector candidate, the candidate determination unit 1420 determines the candidate for the predicted motion vector. At least one predicted motion vector candidate is selectively excluded from the total motion vector candidates by the mathematical formulas (1) to (5), and the predicted motion vector candidate is determined. Of the overall candidates determined based on the motion vector of the block contained in the previously decoded region adjacent to the current block, it is clear that the predictive motion is not used to predict the motion vector of the current block. Vector candidates are excluded based on a given evaluation function.

A virtual motion vector is generated based on information about a predetermined predicted motion vector candidate and a motion vector decoded by the motion vector decoding unit among all candidates of the predicted motion vector, and the generated virtual motion is generated. The virtual motion vector difference, which is the difference between the predicted motion vector candidates different from the vector, is calculated for each of the overall candidates. The calculated virtual motion vector difference is compared with the information about the motion vector decoded by the motion vector decoding unit 1410, that is, the actual motion vector difference, and the candidate of the predicted motion vector is excluded. The entropy coding result of the virtual motion vector difference can be compared with the entropy coding result of the actual motion vector difference, and it can be determined whether or not to exclude the candidate of a predetermined predicted motion vector. Further, in order to improve the estimation accuracy of the entropy coding result, the result of entropy coding of the index information can also be estimated and used for the exclusion judgment. The method of excluding the candidates of the predicted motion vector has been described in relation to the mathematical formulas (1) to (5).

Further, according to still another embodiment of the present invention, the candidate determination unit 1420 rearranges the overall candidates of the predicted motion vector according to a predetermined criterion, and the mathematical formulas (1) to (1) to the rearranged overall candidates. The exclusion determination according to 5) can be repeated to selectively exclude at least one predicted motion vector candidate. It is also possible to exclude duplicate predicted motion vector candidates from the rearranged overall candidates and repeat the exclusion judgment by the formulas (1) to (5).

As a result of the candidate determination unit 1420 excluding at least one predicted motion vector candidate from the overall candidates of the predicted motion vector, if a plurality of predicted motion vector candidates remain among the overall candidates of the predicted motion vector, motion vector decoding is performed. Unit 1410 decodes the information about the predicted motion vector. Information about the predicted motion vector is decoded by a predetermined entropy decoding method. The information about the predicted motion vector is information for identifying the predicted motion vector candidate used for predicting the motion vector of the current block among the predicted motion vector candidates excluding at least one predicted motion vector candidate. Is. Among the candidates for the predicted motion vector not excluded from the candidate determination unit 1420, the information for identifying the predicted motion vector candidate used for predicting the motion vector of the current block is decoded.

As a result of the candidate determination unit 1420 excluding at least one predicted motion vector candidate from the total candidates of the predicted motion vector, if only one predicted motion vector candidate remains, the remaining one predicted motion vector candidate is currently blocked. Since it is used to predict the motion vector, the motion vector decoding unit 1410 does not need to separately decode the information about the candidate of the predicted motion vector.

The motion vector restoration unit 1430 restores the motion vector of the current block based on the information about the motion vector decoded by the motion vector decoding unit 1410. The actual motion vector difference decoded by the motion vector decoding unit 1410 and the predicted motion vector candidate used for predicting the motion vector of the current block are added to restore the motion vector of the current block. Among the predicted motion vector candidates determined by the candidate determination unit 1420, the predicted motion vector candidates used for predicting the motion vector of the current block are determined, and the determined predicted motion vector candidates are referred to as the actual motion vector difference. to add. As a result of exclusion by the candidate determination unit 1420, when a plurality of predicted motion vector candidates other than one are left, the predicted motion vector candidate used for predicting the motion vector of the current block is motion vector decoding. It is determined based on the information about the predicted motion vector decoded in part 1410.

Since the candidate motion vector candidates are determined by the candidate determination unit 1420, the predicted motion vector used to predict the motion vector of the current block even if the information about the decoded predicted motion vector is the same. Candidates may be motion vectors of blocks adjacent to different positions.

FIG. 15 is a flowchart for explaining a method of coding a motion vector according to an embodiment of the present invention. Referring to FIG. 15, at step 1510, the motion vector encoder estimates the motion vector of the current block, and the predicted motion vector candidate used to predict the motion vector of the current block is the predicted motion vector. Decide from all candidates. A motion vector that is the same as or similar to the current block is searched for in a plurality of reference pictures, and the motion vector that is the relative positional difference between the current block and the reference block is estimated from the search results.

It then predicts the motion vector of the current block based on the motion vector of the block contained in the previously encoded region adjacent to the current block. In other words, the motion vector of the block contained in the previously encoded region adjacent to the current block is set as the overall candidate of the predicted motion vector, and among the overall candidates of the predicted motion vector, the estimated current block Determine the predicted motion vector candidates that are most similar to the motion vector. Generates the difference vector between the motion vector of the current block and the determined predicted motion vector candidate, that is, the actual motion vector difference.

At step 1520, the video coding device selectively excludes at least one predicted motion vector candidate from the overall candidates for the predicted motion vector. Of the overall candidates for the predicted motion vector, the predicted motion vector candidates that are clearly not used to predict the motion vector of the current block are excluded.

The video coding apparatus generates a virtual motion vector by using a predetermined predicted motion vector candidate and the actual motion vector difference generated in step 1510 among the overall candidates of the predicted motion vector. The generated virtual motion vector and other predicted motion vector candidates are used to generate a virtual motion vector difference. A virtual motion vector difference can be generated for each of the overall candidates, the generated virtual motion vector difference can be compared with the actual motion vector difference, and a predetermined predicted motion vector candidate can be selectively excluded.

At least one predicted motion vector candidate can be excluded from the overall candidates by iteratively performing the process of generating the virtual motion vector of step 1520 and the process of selectively excluding it for all the overall candidates. .. When the exclusion process is iteratively performed, the virtual motion vector difference is calculated for each of the remaining predicted motion vector candidates excluding the already excluded predicted motion vector candidates, and the calculated virtual motion vector difference is calculated. Can be compared with the actual motion vector difference.

A virtual motion vector difference and an actual motion vector difference can be evaluated and compared based on a predetermined evaluation function, and the predetermined evaluation function is a function for predicting an entropy coding result. A comparison can be made based on a function that estimates the result of entropy-coding the virtual motion vector difference and the result of entropy-coding the actual motion vector difference. In addition, in order to improve the accuracy of the evaluation, the result of entropy-coding the index information can also be estimated and used for determining the exclusion. The method of excluding at least one predicted motion vector candidate from the total candidates of the predicted motion vector has been described in relation to mathematical formulas (1) to (5).

Further, as described in connection with FIGS. 13A to 13D, the motion vector coding apparatus rearranges all the motion vector candidates according to a predetermined reference, and at least one predicted motion vector is selected from the rearranged all candidates. Candidates can also be selectively excluded. It is also possible to exclude duplicate predicted motion vector candidates from the rearranged overall candidates and repeat the exclusion judgment by the formulas (1) to (5).

At step 1530, the motion vector encoder encodes the information about the motion vector and the information about the predicted motion vector. It encodes the actual motion vector difference and the information for identifying the predicted motion vector candidates used to predict the motion vector of the current block. The information about the predicted motion vector is a candidate for the predicted motion vector that has not been excluded through steps 1520 and 1530, and is information for identifying the predicted motion vector candidate used to predict the motion vector of the current block. There may be.

As a result of excluding at least one predicted motion vector candidate from the total candidates of the predicted motion vector, if only one predicted motion vector candidate remains, the information about the predicted motion vector may not be encoded.

FIG. 16 is a flowchart for explaining a method of decoding a motion vector according to an embodiment of the present invention. Referring to FIG. 16, at step 1610, the motion vector decoder decodes information about the motion vector of the current block from the received bitstream. The information about the motion vector may be the difference between the actual motion vector of the current block and the predicted motion vector of the current block.

At step 1620, the motion vector decoder generates a virtual motion vector based on one of the predicted motion vector candidates, the information about the motion vector decoded in step 1610 and the overall candidates for the predicted motion vector. ..

Once the virtual motion vector is generated, the motion vector decoder excludes at least one predicted motion vector candidate from the overall candidates for the predicted motion vector. The overall candidate for the predicted motion vector is determined based on the motion vector of the block in the previously decoded region adjacent to the current block. The motion vector decoding device can selectively exclude at least one predicted motion vector candidate from all such predicted motion vector candidates. Based on a predetermined evaluation function, the virtual motion vector difference and the actual motion vector difference decoded in step 1610 are evaluated, and the predetermined predicted motion vector candidate is selectively excluded. The method of excluding the predicted motion vector candidates from the overall candidates is the same as in step 1530, and has been described with reference to mathematical formulas (1) to (5).

At least one predicted motion vector candidate can be excluded from the overall candidates by iteratively performing the process of generating the virtual motion vector of step 1620 and the process of selectively excluding it for all the overall candidates. can.

Further, as described in connection with FIGS. 13A to 13D, the motion vector coding apparatus rearranges all the motion vector candidates according to a predetermined reference, and at least one predicted motion vector is selected from the rearranged all candidates. Candidates can also be selectively excluded. It is also possible to exclude duplicate predicted motion vector candidates from the rearranged overall candidates and repeat the exclusion judgment by the formulas (1) to (5).

As a result of exclusion, if a plurality of predicted motion vector candidates remain, the information about the predicted motion vector is decoded, and if only one predicted motion vector candidate remains, the information about the predicted motion vector is not decoded.

At step 1630, the motion vector decoder determines the predicted motion vector candidates used to predict the motion vector of the current block, among the predicted motion vector candidates not excluded from step 1620.

Based on the information about the predicted motion vector of the current block, among the candidates of the predicted motion vector, the predicted motion vector candidate used for predicting the motion vector of the current block can be determined. As a result of the exclusion in step 1620, if only one predicted motion vector candidate remains, the remaining one predicted motion vector candidate is determined as the predicted motion vector candidate used to predict the motion vector of the current block. ..

When the predicted motion vector candidate is determined, the determined predicted motion vector candidate and the actual motion vector difference decoded in step 1610 are added to restore the motion vector of the current block.

As described above, even if the present invention is described with reference to limited embodiments and drawings, the present invention is not limited to the above-described embodiments, and those skilled in the art in the field to which the present invention belongs. If so, various modifications and modifications would be possible from such a description. Therefore, the idea of the present invention must be grasped only by the scope of claims, and any modification equal to or equivalent to this belongs to the category of the idea of the present invention. Further, the system according to the present invention can be embodied as a computer-readable code on a computer-readable recording medium.

For example, the video coding device, the video decoding device, the motion vector coding device, and the motion vector decoding device according to the exemplary embodiment of the present invention are shown in FIGS. 1, 2, 4, 5, 9, and 9. A bus coupled to each unit of the device as illustrated in 14, may include at least one processor coupled to the bus. It may also include memory coupled to said bus and coupled to at least one processor for carrying out instructions as described above in order to store instructions, received messages or generated messages.

Also, computer-readable recording media include all types of recording devices in which data readable by computer systems is stored. Examples of recording media include ROM (read-only memory), RAM (random-access memory), CD-ROM, magnetic tape, floppy (registered trademark) disk, optical data storage device, and the like. In addition, computer-readable recording media are distributed to computer systems connected to a network, and computer-readable code is stored and executed in a distributed manner.

1400 Vector decoding device 1410 Motion vector decoding unit 1420 Candidate determination unit 1430 Motion vector restoration unit

Claims (10)

In the method of decoding video
The stage of decoding the information about the motion vector difference of the current block in the video and the information about the predicted motion vector for specifying the predicted motion vector of the current block from the bit stream, and
The stage of constructing the predicted motion vector candidate set and
A step of adjusting the predicted motion vector candidate set based on the value of the predicted motion vector candidate in the predicted motion vector candidate set, and a step of adjusting the predicted motion vector candidate set.
A step of determining the predicted motion vector of the current block based on the adjusted predicted motion vector candidate set and the information about the predicted motion vector, and
Including a step of determining the motion vector of the current block based on the information about the predicted motion vector of the current block and the motion vector difference of the current block.
The adjusted predicted motion vector candidate set is based on the first-order predicted motion vector candidate based on the motion vector of the block adjacent to the current block and the motion vector of the block on the reference picture at the same position as the current block. Includes at least one of the next predicted motion vector candidates
The adjacent block includes a first block at the lower left of the current block and a second block at the top of the first block.
The video is divided into a plurality of maximum coding units according to the maximum coding unit size information, and the maximum coding unit is hierarchically divided by one or more coding units having a depth to encode the current depth. The unit is one of the coding units divided from the higher depth coding unit.
A decoding method, wherein the current block is included in the coding unit of the current depth. The step of adjusting the predicted motion vector candidate set is
When there are primary predicted motion vector candidates having overlapping values in the predicted motion vector candidate set, one of the primary predicted motion vector candidates having overlapping values is removed by the predicted motion vector candidate set. The decoding method according to claim 1, wherein the decoding method is characterized by the above. The adjusted predicted motion vector candidate set contains a total of two predicted motion vector candidates.
The decoding method according to claim 1, wherein the predicted motion vector information indicates one of the two predicted motion vector candidates through a 1-bit binary numerical value. A decoding unit that decodes information about the motion vector difference of the current block in the video and information about the predicted motion vector for specifying the predicted motion vector of the current block from the bit stream.
A candidate determination unit that constitutes a predicted motion vector candidate set and adjusts the predicted motion vector candidate set based on the values of the predicted motion vector candidates in the predicted motion vector candidate set.
Based on the information about the adjusted predicted motion vector candidate set and the predicted motion vector, the predicted motion vector of the current block is determined, and the information about the predicted motion vector of the current block and the motion vector difference of the current block is used. Based on, the motion vector restoration unit that determines the motion vector of the current block is included.
The adjusted predicted motion vector candidate set is based on the first-order predicted motion vector candidate based on the motion vector of the block adjacent to the current block and the motion vector of the block on the reference picture at the same position as the current block. Includes at least one of the next predicted motion vector candidates
The adjacent block includes a first block at the lower left of the current block and a second block at the top of the first block.
The video is divided into a plurality of maximum coding units according to the maximum coding unit size information, and the maximum coding unit is hierarchically divided by one or more coding units having a depth to encode the current depth. The unit is one of the coding units divided from the higher depth coding unit.
A decoding device, characterized in that the current block is included in the coding unit of the current depth. When adjusting the predicted motion vector candidate set, the candidate determination unit determines that if there are primary predicted motion vector candidates having overlapping values in the predicted motion vector candidate set, the primary predicted motion having the overlapping values is present. The decoding device according to claim 4, wherein one of the vector candidates is removed by the predicted motion vector candidate set. The candidate determination unit adjusts the predicted motion vector candidate set so as to include a total of two predicted motion vector candidates.
The decoding device according to claim 4, wherein the candidate determination unit determines one of the two predicted motion vector candidates through the predicted motion vector information having a 1-bit binary value. In the method of coding video
At the stage of performing inter-prediction for the current block in the video and determining the motion vector of the current block,
The stage of constructing the predicted motion vector candidate set and
A step of adjusting the predicted motion vector candidate set based on the value of the predicted motion vector candidate in the predicted motion vector candidate set, and a step of adjusting the predicted motion vector candidate set.
Among the predicted motion vector candidates included in the adjusted predicted motion vector candidate set based on the motion vector of the current block, the predicted motion vector of the current block is determined, and the predicted motion vector of the current block is specified. And the stage of encoding information about the predicted motion vector for
A step of encoding information about the difference between the motion vector of the current block and the predicted motion vector of the current block, which indicates the difference between the motion vector of the current block and the predicted motion vector of the current block.
Including a step of generating a bitstream containing information about the motion vector difference of the current block and information about the predicted motion vector for identifying the predicted motion vector of the current block.
The adjusted predicted motion vector candidate set is based on the first-order predicted motion vector candidate based on the motion vector of the block adjacent to the current block and the motion vector of the block on the reference picture at the same position as the current block. Includes at least one of the next predicted motion vector candidates
The adjacent block includes a first block at the lower left of the current block and a second block at the top of the first block.
The video is divided into a plurality of maximum coding units according to the maximum coding unit size, and the maximum coding unit is hierarchically divided by one or more coding units having a depth, and the coding unit of the current depth. Is one of the coding units divided from the higher depth coding unit,
A coding method, wherein the current block is included in the coding unit of the current depth. In a device that encodes video
A motion vector estimation unit that performs inter-prediction for the current block in the video and determines the motion vector of the current block.
A candidate determination unit that constitutes a predicted motion vector candidate set and adjusts the predicted motion vector candidate set based on the values of the predicted motion vector candidates in the predicted motion vector candidate set.
To determine the predicted motion vector of the current block among the predicted motion vector candidates included in the adjusted predicted motion vector candidate set based on the motion vector of the current block, and to specify the predicted motion vector of the current block. The information about the predicted motion vector of the current block is encoded, and the information about the difference of the motion vector of the current block indicating the difference between the motion vector of the current block and the predicted motion vector of the current block is encoded, and the motion of the current block is encoded. Includes a motion vector encoding unit that generates a bit stream containing information about the vector difference and information about the predicted motion vector for identifying the predicted motion vector of the current block.
The adjusted predicted motion vector candidate set is based on the first-order predicted motion vector candidate based on the motion vector of the block adjacent to the current block and the motion vector of the block on the reference picture at the same position as the current block. Includes at least one of the next predicted motion vector candidates
The adjacent block includes a first block at the lower left of the current block and a second block at the top of the first block.
The video is divided into a plurality of maximum coding units according to the maximum coding unit size, and the maximum coding unit is hierarchically divided by one or more coding units having a depth, and the coding unit of the current depth. Is one of the coding units divided from the higher depth coding unit,
A coding device, characterized in that the current block is included in the coding unit of the current depth. In the method of coding video
At the stage of performing inter-prediction for the current block and determining the motion vector of the current block,
The stage of constructing the predicted motion vector candidate set and
A step of adjusting the predicted motion vector candidate set based on the value of the predicted motion vector candidate in the predicted motion vector candidate set, and a step of adjusting the predicted motion vector candidate set.
To determine the predicted motion vector of the current block among the predicted motion vector candidates included in the adjusted predicted motion vector candidate set based on the motion vector of the current block, and to specify the predicted motion vector of the current block. And the stage of encoding information about the predicted motion vector of
A step of encoding information about the difference in the motion vector of the current block, which indicates the difference between the motion vector of the current block and the predicted motion vector of the current block.
Including a step of generating a bitstream containing information about the motion vector difference of the current block and information about the predicted motion vector for identifying the predicted motion vector of the current block.
The adjusted predicted motion vector candidate set is based on the first-order predicted motion vector candidate based on the motion vector of the block adjacent to the current block and the motion vector of the block on the reference picture at the same position as the current block. Includes at least one of the next predicted motion vector candidates
A coding method, wherein the adjacent block includes a first block at the lower left of the current block and a second block at the top of the first block. In a device that encodes video
A motion vector estimation unit that performs inter-prediction for the current block and determines the motion vector of the current block.
A candidate determination unit that constitutes a predicted motion vector candidate set and adjusts the predicted motion vector candidate set based on the values of the predicted motion vector candidates in the predicted motion vector candidate set.
To determine the predicted motion vector of the current block among the predicted motion vector candidates included in the adjusted predicted motion vector candidate set based on the motion vector of the current block, and to specify the predicted motion vector of the current block. The information about the predicted motion vector of the current block is encoded, and the information about the difference of the motion vector of the current block indicating the difference between the motion vector of the current block and the predicted motion vector of the current block is encoded, and the motion of the current block is encoded. Includes a motion vector encoding unit that generates a bit stream containing information about the vector difference and information about the predicted motion vector for identifying the predicted motion vector of the current block.
The adjusted predicted motion vector candidate set is based on the first-order predicted motion vector candidate based on the motion vector of the block adjacent to the current block and the motion vector of the block on the reference picture at the same position as the current block. Includes at least one of the next predicted motion vector candidates
The encoding device, wherein the adjacent block includes a first block at the lower left of the current block and a second block above the first block.

Images