JP3092281B2 - Highly efficient encoding and decoding apparatus for image signals - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a high-efficiency image signal encoding apparatus for encoding an image signal by orthogonal transformation with high efficiency, and a decoding apparatus therefor.

[0002]

2. Description of the Related Art As a method of encoding an image signal with high efficiency, for example, MPEG (Moving Picture Experts Group) is used.
In the standardization proposal according to p), a high-efficiency encoding method for image signals for so-called digital storage media is specified. Here, the storage media targeted by this method is a so-called CD (compact disk) or D
It has a continuous transfer speed of about 1.5 Mbit / sec or less, such as an AT (digital audio tape) or a hard disk. It is also assumed that this is connected not only directly to the decoder but also via a transmission medium such as a computer bus, a LAN (local area network), and telecommunications.
Further, special functions such as random access, high-speed reproduction, reverse-order reproduction, and the like are considered in addition to normal-order reproduction.

The principle of the high-efficiency encoding method of the image signal by the MPEG is as follows.

That is, in this high-efficiency coding method, first, a difference between images is obtained to reduce redundancy in a time axis direction, and thereafter, a so-called discrete cosine transform (DCT) process and a variable length code are used. To reduce the redundancy in the space axis direction.

First, the redundancy in the time axis direction will be described below.

In general, in a continuous moving image, an image before and after in time is similar to an image of interest (ie, an image at a certain time). For this reason, as shown in FIG. 22, for example, if a difference between an image to be encoded and a temporally forward image is obtained and the difference is transmitted, the redundancy in the time axis direction can be reduced. It is possible to reduce the amount of information to be transmitted. The image encoded in this manner is a forward predictive-coded picture (P-picture or P-picture or P-picture) described later.
Frame). Similarly, the difference between the image to be encoded and the interpolated image formed from the front or the rear or the front and the rear in time is calculated, and the difference between the small values is transmitted. For example, it is possible to reduce the amount of information to be transmitted by reducing the redundancy in the time axis direction. An image encoded in this way is a bidirectional predictive encoded image (Bidirectio
nallyPredictive-coded picture, B picture or B
Frame). In FIG. 22, an image indicated by I in the figure indicates an intra-coded image (Intra-coded picture, I picture or I frame) described later, and an image indicated by P in the figure indicates the P A picture is shown, and an image shown by B in the figure shows the B picture.

In order to generate each predicted image, so-called motion compensation is performed. That is, according to this motion compensation, for example, a block of, for example, 16 × 16 pixels (hereinafter, referred to as a macroblock) constituted by a unit block of 8 × 8 pixels is created, and the block is located near the position of the macroblock in the previous image. By searching for a place with a small difference and calculating the difference with the searched macroblock, data to be sent can be reduced. Actually, for example, in the P picture (forward prediction coded image), the difference between the prediction image after motion compensation and the difference between the prediction image after motion compensation and Those with a small amount are selected and coded in macroblock units of 16 × 16 pixels.

However, in the case described above, a large amount of data must be sent, for example, with respect to a portion (image) coming out from behind the moving object. Therefore, for example, in the B picture (bidirectionally coded image), the already decoded temporally forward or backward image after motion compensation, and the interpolated image formed by adding both of them, will be coded from now on. And the image which does not take the difference, that is, the image having the smallest data amount among the four images to be encoded from now.

Next, the redundancy in the space axis direction will be described below.

The difference between the image data is not transmitted as it is, but is subjected to discrete cosine transform (DCT) for each unit block of 8 × 8 pixels. The DCT expresses an image not by a pixel level but by how many frequency components of a cosine function are included. For example, by a two-dimensional DCT, data of a unit block of 8 × 8 pixels is two-dimensional. It is converted into a coefficient block of an 8 × 8 cosine function component by DCT. For example, an image signal of a natural image taken by a television camera is often a smooth signal. In this case, the DCT is applied to the image signal.
By performing the processing, the data amount can be efficiently reduced.

That is, for example, in the case of a smooth signal such as the above-described image signal of a natural image, a large value concentrates around a certain coefficient by applying the DCT. When these coefficients are quantized, most of the 8 × 8 coefficient blocks become 0, and only large coefficients remain. Therefore, when transmitting the data of the 8 × 8 coefficient block, a set of non-zero coefficients and a so-called zero run indicating how much 0 has continued before the coefficient in a so-called zigzag scan order. By transmitting the so-called Huffman code, the amount of transmission can be reduced. On the decoder side, an image is reconstructed in the reverse procedure.

FIG. 23 shows the structure of data handled by the above-described encoding method. That is, the data structure shown in FIG. 23 includes, in order from the bottom, a block layer, a macroblock layer, a slice layer, a picture layer, a group of picture (GOP: Group of Picture) layer, and a video sequence layer. . Hereinafter, description will be made in order from the lower layer in FIG.

First, in the above-mentioned block layer, a block of the block layer is composed of 8 × 8 pixels (8 lines × 8 pixels) having adjacent luminance or color difference. The DCT (discrete cosine transform) described above is applied to each unit block.

In the macroblock layer, the macroblocks in the macroblock layer are composed of four luminance blocks (luminance unit blocks) Y0,
Y1, Y2, Y3 and a color difference block (color difference unit block) C at the same position as the luminance block on the image.
r and Cb are composed of a total of six blocks. The order of transmission of these blocks is Y0, Y1, Y2, Y3, C
r, then Cb. Here, in the encoding method, what is used for a predicted image (a reference image for obtaining a difference)
Alternatively, whether or not the difference need not be transmitted is determined in units of this macroblock.

The slice layer is composed of one or a plurality of macroblocks connected in the scanning order of the image. At the beginning of the slice (header), the difference between the motion vector and the DC (direct current) component in the image is reset, and the first macroblock has data indicating the position in the image, so that an error occurs. It is designed to allow you to return if something happens. Therefore, the length of the slice and the starting position are arbitrary, and can be changed according to the error state of the transmission path.

In the picture layer, a picture, that is, an image one by one, is composed of at least one or a plurality of the slices. Then, each of the above-described intra-coded images (I
Picture or I frame), forward predictive coded picture (P picture or P frame), bidirectional predictive coded picture (B picture or B frame), DC intra coded picture (DC coded (D) picture) Classified into images.

Here, in the above-mentioned intra-coded image (I picture), at the time of coding, only closed information in one image is used. Therefore, in other words, when decoding, an image can be reconstructed using only the information of the I picture itself. Actually, encoding is performed by DCT processing without taking a difference. This encoding method is generally inefficient, but if it is included everywhere, random access and high-speed reproduction can be performed.

In the forward predictive coded picture (P picture), an I picture or P picture which is located earlier in time at the input and has already been decoded is used as a predictive picture (picture serving as a reference for obtaining a difference). I do. In practice, the more efficient one of encoding the difference from the motion-compensated predicted image and encoding the difference without taking the difference (intra code) is selected in units of the macroblock.

The above bidirectional predictive coded image (B picture)
Uses three types of predicted images, i.e., an I-picture or a P-picture that is located earlier in time and has already been decoded, and an interpolated image created from both of them. This allows
The most efficient one of the above three types of difference-encoded differential coding and intra-coding can be selected in macroblock units.

The DC intra-coded image is an intra-coded image composed of only DCT DC coefficients.
It cannot exist in the same sequence as the other three images.

The group of pictures (GOP) layer includes one or more I-pictures and zero or more non-I-pictures.
And a picture. Here, the input order to the encoder is, for example, 1I, 2B, 3B, 4P * 5B, 6
B, 7I, 8B, 9B, 10I, 11B, 12B, 13
P, 14B, 15B, 16P * 17B, 18B, 19
When I, 20B, 21B, and 22P are used, the output of the encoder, that is, the input of the decoder is, for example, 1I, 4
P, 2B, 3B * 7I, 5B, 6B, 10I, 8B, 9
B, 13P, 11B, 12B, 16P, 14B, 15B
* 19I, 17B, 18B, 22P, 20B, 21B. The reason why the rearrangement of the order is performed in the encoder is, for example, in the case of encoding or decoding the B picture, the I picture or the P picture which is a temporally backward prediction image. This is because the picture must be coded first. Here, the interval (for example, 9) of the I picture and the interval (for example, 3) of the I picture or the B picture are free. Also, I picture or P
The picture interval may vary within the group of pictures layer. The breaks in the group of picture layers are indicated by *. In addition, I indicates an I picture, P indicates a P picture, and B indicates a B picture.

The video sequence layer has an image size,
It is composed of one or more group of picture layers having the same image rate.

[0023]

As described above, when transmitting a moving image standardized by the above-described high-efficiency coding method by MPEG, an image obtained by compressing one image in a picture is transmitted first. Then, the difference between this image and the image whose motion has been compensated is transmitted.

However, in the above-mentioned one image, for example, when processing a field as a picture, 2
Since the vertical position will be different in the field alternately,
For example, when transmitting a still image, difference information must be transmitted.

When a frame is processed as a picture, for example, a moving part in the frame must be processed as a so-called comb-shaped image. That is, for example, as shown in FIG. 24, when there is a moving object CA such as a car in front of a stationary background, when one frame is viewed, there is movement between fields. turn into.

Further, for example, when processing an image in which a still portion and a moving image portion are mixed, no matter whether the above-mentioned field is processed as a picture or the frame is processed as a picture, In this case, an image of a part having a low compression efficiency is formed.

Accordingly, the present invention has been proposed in view of the above-described circumstances. For a moving image having a field structure, an image having little motion or an image having much motion, or an image in which both of them are mixed, is used. It is an object of the present invention to provide a high-efficiency image signal encoding apparatus and a decoding apparatus capable of efficiently performing field processing or frame processing.

[0028]

In order to solve such a problem, a high-efficiency coding apparatus according to the present invention provides a high-efficiency coding apparatus for coding an image signal to be coded in units of a macroblock consisting of a two-dimensional array of a plurality of pixels. In the high-efficiency coding apparatus, a means for detecting a motion vector between frames and a sum of absolute value differences of each pixel in the macroblock unit and an odd or even number of scans of pixels in the frame in the macroblock unit are used. Motion detecting means comprising a motion vector between fields and means for detecting the sum of absolute value differences of each pixel; a frame prediction mode for performing motion compensation in units of frames in the macroblock; and a unit in fields in the macroblock. Which of the field prediction modes that perform motion compensation is more efficient in performing motion compensation? First mode selection means for selecting an efficient prediction mode, and a frame processing mode for performing orthogonal transform on a frame-by-frame basis in the macroblock and an orthogonal transform on a field in the macroblock basis. It is determined which of the field processing modes in which the block is formed so as to perform the orthogonal transformation is more efficient when performing the orthogonal transformation, by using information output from the motion detecting means and the first mode selecting means. Second mode selecting means for selecting a mode of blocking, and for each macroblock in one frame, the blocking is adaptively switched to the frame processing mode or the field processing mode, and each macroblock is set based on each mode. Encoding mode that encodes all macros in one frame The blocking of the lock is performed in the field processing mode, and the odd cycle of the scanning of the odd field in the interlace encodes only one frame of the odd field in the macroblock, and then the scanning of the even field in the interlace. It is determined which of the second encoding processing modes for encoding even frames in the macroblock for one frame in the even-numbered cycle is more efficient at the time of encoding, and the more efficient encoding processing mode is selected. The third mode selecting means, and recognizes whether the odd cycle or the even cycle, and if the encoding processing mode is the first encoding processing mode, blocks in the odd cycle corresponding to the blocking mode. Memory group to output the macro block When the encoding mode is the second encoding mode, the frame memory group is configured to output a macroblock that is divided into blocks corresponding to the field processing mode in the odd cycle and the even cycle. Receiving the motion prediction mode information selected by the first mode selection means and the blocking mode information selected by the second mode selection means, and controlling the motion according to the mode information. And a motion compensating means for executing a compensation frame or inter-field prediction.

In other words, the high-efficiency encoding apparatus of the present invention, in encoding a moving image in which one frame is composed of two fields, has an odd field (first field) and an even field (first field) for all blocks in the frame. An encoding means (second encoding processing mode) that divides the second field) into blocks and enables motion prediction from the first field to the second field;
Coding means (first coding processing mode) for switching between dividing / not dividing the second field on a macroblock basis to form a block, and these coding means are adaptively switched for each frame. It was made.
Further, in this case, 1-bit information indicating the encoding means (information indicating the selected mode) is added to the code.

Further, the high-efficiency decoding apparatus of the present invention receives reproduced image coded data and header information including detected motion vector information, motion prediction mode information, blocking mode information, and coding processing mode information. Inverse variable-length encoding means for outputting the detected motion vector information, the motion prediction mode information, the blocking mode information, and the encoding mode information together with the decoded image decoded data; Address generation means for calculating an address increment value in a frame buffer for storing the image decoding data from the conversion processing mode information, obtaining a head address of each macroblock, and giving the head address to the frame buffer; The relative address of the macro block other than the start address is added to the frame buffer to And receives the detected motion vector information, the motion prediction mode information, the blocking mode information, and the encoding processing mode information, performs motion compensation corresponding to the mode information, and obtains a motion-compensated image signal. Is provided to the frame buffer.

[0031]

According to the present invention, the first encoding mode and the second encoding mode can be switched on a macroblock basis. In the first encoding processing mode, efficient encoding can be performed by adaptively selecting the frame processing mode and the field processing mode according to, for example, the magnitude of the motion of the image. In the second encoding processing mode, by setting all the macroblocks in the frame to be in the field processing mode, for example, it becomes possible to perform motion prediction from an odd field to an even field for a frame having particularly large motion. Moving images can be encoded well.

[0032]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 shows a high-efficiency image signal encoding apparatus according to a first embodiment of the present invention. This is a high-efficiency encoding apparatus for encoding an image signal, which performs encoding in units of a block of 16 × 16 pixels in a spatial arrangement of input image data.

The high-efficiency image signal encoding apparatus according to the first embodiment shown in FIG. 1 has a frame (one screen) composed of a plurality of unit blocks (macroblocks) of 16 × 16 pixels. A frame memory group 10 that is stored as a plurality of original images, a frame motion detecting circuit 22 which is a means for detecting a motion vector between frames and an absolute value difference sum of each pixel in the macroblock unit, and It has a motion detecting means comprising a motion vector between fields, which are obtained by dividing the scan of the pixels of the frame by odd or even numbers, and a field motion detecting circuit 21 which detects the sum of absolute value differences of each pixel.

The apparatus according to the present embodiment includes a frame prediction mode (hereinafter, referred to as a frame motion prediction mode) for performing motion compensation in units of frames in the macroblock and a field prediction mode for performing motion compensation in units of fields in the macroblock. A mode (hereinafter referred to as a field motion prediction mode) to determine which one is more efficient in performing motion compensation, and select a more efficient motion prediction mode; A frame processing mode (hereinafter referred to as a frame orthogonal transform mode) in which blocks are formed so as to perform orthogonal transform (for example, discrete cosine transform; DCT) as a unit.
And the field processing mode (hereinafter, referred to as a field orthogonal transformation mode) that performs blocking so that orthogonal transformation is performed in units of fields in the macroblock. There is provided a frame / field mode determination circuit 33 including a second mode selection unit that determines using the information output from the first mode selection unit and selects an efficient blocking mode.

Further, the apparatus according to the present embodiment, together with the motion detecting means and the frame / field mode determination circuit 33, converts the orthogonal transform into blocks in the frame orthogonal transform mode or the field orthogonal transform for each macroblock in one frame. The first encoding processing mode in which each macroblock is encoded based on each mode by adaptively switching to the transform mode, and the orthogonal transform blocking of all the macroblocks in one frame is performed in the field orthogonal transform mode. In the odd cycle (odd cycle) of the scanning period of the odd field in the interlace, only the odd field of the macroblock is encoded for one frame, and the even cycle (even cycle) of the scanning period of the even field in the interlace is performed.
A third mode for determining which of the second encoding processing modes for encoding even fields in a macroblock for one frame is more efficient in encoding, and selecting an efficient encoding processing mode It has an encoding processing mode determination circuit 34 as selection means.

Further, the apparatus of the present embodiment recognizes whether the cycle is the odd cycle or the even cycle. The frame memory group 10 is controlled so as to output a macroblock which has been divided into blocks corresponding to the encoding mode. And an address generator 35 for controlling the frame memory group 10 so as to output a macroblock divided into blocks corresponding to the field orthogonal transform mode, and the motion prediction mode information selected by the first mode selecting means. (Frame / field motion prediction mode data) and the blocking mode information ( Frame / field orthogonal transform mode data) it receives, in which a motion compensator with a frame memory group 20 is a motion compensation means in response to these mode information to perform the motion compensation frame or inter-field prediction.

First, prior to the description of the specific configuration of the apparatus of the first embodiment of the present invention shown in FIG. 1, a high-efficiency encoding process of an image signal performed by the apparatus of the present embodiment will be described.

In the encoding apparatus of this embodiment, as shown in FIG. 2, for example, intra-frame encoding (I frame or I frame
Picture), unidirectional prediction interframe coding (P frame or P picture), and bidirectional inter picture coding (B frame or B picture). Each picture is divided into 8 × 8 pixels and is divided into 2 × 2 blocks (ie, 16 × 16 pixels).
Form a macroblock.

Here, in the coding apparatus of this embodiment, the first mode selection means selects which of the above-described frame motion prediction mode and field motion prediction mode is more efficient in performing motion compensation. The second mode selection means can select which of the frame orthogonal transform mode and the field orthogonal transform mode is more efficient in performing the orthogonal transform. In addition,
The first and second mode selections are made by the frame / field mode determination circuit 33 described above.

In addition, the apparatus of the present embodiment performs the encoding in each of the two encoding processing modes, whichever is more efficient, for each frame together with the mode selection processing by the first and second mode selection means. It has been done. That is, as described above, as the first encoding processing mode, each macroblock is encoded by adaptively switching between the frame orthogonal transform mode and the field orthogonal transform mode for each macroblock in one frame. . Further, as described above, as the second encoding processing mode, the blocking of the orthogonal transform of all the macroblocks in one frame is performed in the field orthogonal transform mode, and the odd field (first
Field), the odd-numbered cycle of the scan period of the field) is encoded for one frame only in the odd-numbered field of the macroblock, and the even-numbered field of the macroblock is encoded for one frame in the even-numbered cycle of the scan period of the even-numbered field (second field). Become The third mode selecting means determines which of the first and second encoding processing modes is more efficient for encoding, and selects an efficient encoding processing mode.

That is, in the first encoding processing mode, each frame is divided into a first field (odd field) and a second field (even field) without being divided into blocks and encoded (the frame orthogonal transform). Mode) and a mode in which each frame is divided into first and second fields and divided into fields and encoded (the above-described field orthogonal transform mode). In contrast, for example, a macroblock having a large image motion is adaptively switched to use the field orthogonal transform mode.

Therefore, when the frame orthogonal transform mode is selected in the first encoding processing mode,
For example, in the motion prediction of the P and B frames, motion prediction is performed from the previous and subsequent frames, and a difference image from the predicted image is subjected to orthogonal transform (DCT). Also, when the field orthogonal transform mode is selected in the first encoding processing mode, for example, the motion prediction of the P and B frames is performed for the first field and the second field of the first and second fields of the macroblock. Motion is predicted from the first or second field of the frame, and a difference image from the predicted image is a DCT.
Is converted. Thus, it can be said that the first encoding processing mode is encoding without intra-frame inter-field prediction. In the first encoding mode, encoding is performed in the odd cycle. In addition,
This first encoding processing mode can be said to be encoding without intra-frame inter-field prediction.

Here, in the first encoding mode, it is impossible to predict the motion between the fields in the frame (between the odd field and the even field in the same frame).

Therefore, in the second encoding processing mode of the present embodiment, as described above, all macroblocks in each frame are subjected to orthogonal transform blocking in the field orthogonal transform mode. In the odd cycle, only the odd field in the macro block is encoded for one frame, and then in the even cycle, the even field in the macro block is encoded for one frame. Therefore, according to the second encoding mode, since the odd field (first field) is encoded first, the even field (second field) is subjected to motion prediction from this odd field (first field). Becomes possible. From this, it can be said that the second encoding mode is encoding of a frame with intra-frame inter-field prediction.

Returning to FIG. 1 again, the main flow of image data to be coded by the coding apparatus according to the present embodiment will be described using the configuration of FIG.

That is, in FIG.
Is supplied with a digital image signal and stored in the frame memory group 10. From the frame memory group 10, data of the unit macroblock of 16 × 16 pixels is controlled and read by an address generator 35 described later, and is transmitted to the difference detector 12. The difference detector 12 includes a frame memory group 20 with a motion compensator described later.
Are also supplied, and the difference detector 12 detects these differences.

The output of the difference detector 12 is sent to a DCT circuit 13 that performs an orthogonal transform (DCT) process. The DCT coefficient data obtained by the DCT processing by the DCT circuit 13 is sent to the quantizer 14. The quantizer 14
Is output from the output terminal 2 as encoded data via a variable-length encoding circuit 15 and a buffer 16, which perform variable-length encoding such as Huffman encoding or run-length encoding. .

In the frame memory group with motion compensator 20, the quantized data from the quantizer 14 is an inverse quantizer that performs an inverse quantization process of the quantization process in the quantizer 14. Data is supplied via an adder 19 through an inverter 17 and an inverse DCT circuit 18 for performing an inverse DCT process of the DCT process in the DCT circuit 13.
In the adder 19, the output of the inverse DCT circuit 18 and the output of the frame memory group with motion compensator 20 are added. Note that the buffer 1
The information for preventing the overflow of No. 6 is fed back to the quantizer 14.

On the other hand, the image data output from the frame memory group 10 in macroblock units is transmitted to a frame motion detection circuit 22 and a field motion detection circuit 21.

The frame motion detecting circuit 22 detects the inter-frame motion vector and the sum of absolute differences of each pixel in units of the macroblock, and outputs these data (the data of the inter-frame motion vector FMMV and the sum of absolute differences). Data FMAD) is output. Further, the field motion detection circuit 21 detects a motion vector between fields and the sum of absolute value differences of each pixel in the macroblock unit, and outputs these data (data FDMV of the motion vector between fields and sum of absolute value differences). Data FDAD). The data FMMV of each motion vector of these motion detection circuits 21 and 22
/ FDMV is transmitted to the selector 24.

The data FMAD / FDAD of the sum of absolute difference from the field motion detection circuit 21 and the frame motion detection circuit 22 and the data FMMV /
The FDMV is the frame / field mode determination circuit 3
Also sent to 3.

The frame / field mode determination circuit 33 performs the following motion based on the absolute value difference sum data FMAD from the frame motion detection circuit 22 and the absolute value difference sum data FDAD from the field motion detection circuit 21. At the time of the motion prediction processing in the frame memory group with compensator 20, it is determined whether the motion prediction processing is performed in the frame unit or the field prediction unit. ) Is output. More specifically, in the frame / field mode determination circuit 33, for example, the difference between the absolute value difference sum data FMAD and the absolute value difference sum data FDAD is larger than a certain threshold value T1 (FMAD-FDAD> T1). Hour),
The circuit 33 outputs data (data MPFD of the field motion prediction mode in the motion prediction) indicating that it is more efficient to perform the motion prediction processing on a field-by-field basis. Conversely, if it is determined that the difference between the absolute value difference sum data FMAD and the absolute value difference sum data FDAD is smaller than or equal to the threshold value T1 (when FMAD-FDAD ≦ T1), the frame unit And outputs data (frame motion prediction mode data MPFM in motion prediction) indicating that the motion prediction process is more efficient.

Any of these motion prediction mode data MP
FM / MPFD is the frame memory group with motion compensator 20
Sent to In addition, these motion prediction mode data MPFM
/ MPFD is also sent to the selector 24.

In response to the motion prediction mode data MPFM / MPFD from the frame / field mode determination circuit 33, the selector 24 controls the frame motion detection circuit 2
2 and the inter-field motion vector data FDMV supplied from the field motion detection circuit 21.
Selective output. That is, the motion prediction mode data is data M indicating the field motion prediction mode.
At the time of PFD, the motion vector data FDMV from the field motion detection circuit 21 is selected and output, and the motion prediction mode data is data M indicating the frame motion prediction mode.
At the time of PFM, the motion vector data FMMV from the frame motion detection circuit 22 is selected and output. The motion vector data FMMV / FDMV selected by the selector 24
Are sent to the frame memory group with motion compensator 20. Thus, the frame memory group 20 can perform motion compensation on a frame basis or on a field basis based on the motion prediction mode data MPFM / MPFD and the motion vector data FMMV / FDMV.

Further, the frame / field mode determination circuit 33 is also supplied with the image data in macroblock units read from the frame memory group 10. The frame / field mode determination circuit 33
Then, a difference image is created using the motion prediction mode data MPFM / MPFD, the motion vector data FMMV / FDMV, and the image data from the frame memory group 10, and based on the difference image, the difference image is generated.
The processing for selecting the orthogonal transformation block processing mode (the frame orthogonal transformation mode / field orthogonal transformation mode) most suitable for the image output from 0 and subjected to the DCT processing by the DCT circuit 13 is also performed at the same time. ing. In the case of the I picture (or I frame), data of an image (original image) in the frame memory group 10 is used instead of the difference image.

That is, it is assumed here that, for example, the macroblock in the difference image is a macroblock as shown in FIG. 3 (an original picture macroblock in an I picture). In FIG. 3, the odd lines (o1, o)
2, o3,... ON, where N is a macroblock 16) is indicated by a solid line, and even lines (e1, e2, e3)
,... EN, where N is 1 for a macroblock
6) is indicated by a dotted line. Each pixel of the even line is expressed as e (i, j), and each pixel o (i,
j). In the difference image or the original image (I-picture image) as shown in FIG. 3, the difference EFD of the difference image in the field unit can be expressed by the following mathematical expression 1. The difference EFM of the difference image in the frame unit is It can be shown by the following mathematical expression 2.

[0058]

(Equation 1)

[0059]

(Equation 2)

In the frame / field mode determination circuit 33, specifically, a threshold value having a difference between the difference EFM obtained in the frame and the difference EFD obtained in the field by using the formulas 1 and 2 is used. Greater than T2 (EFM
If (−FDD> T2), it outputs data (data MDFD in the field orthogonal transform mode in the orthogonal transform blocking process) indicating that the DCT in the DCT circuit 13 is to be performed on a field basis. Conversely, when it is determined that the difference between the difference EFM and the difference EFD is smaller than or equal to the threshold value T2 (when EFM-EFD≤T2), the DCT in the DCT circuit 13 is set to the above value. Data indicating that the processing is performed in frame units (data M in the frame orthogonal transformation mode in the orthogonal transformation block processing).
DFM).

Here, the frame orthogonal transformation mode data MD from the frame / field mode decision circuit 33 is output.
The output of the FM or field orthogonal transform mode data MDFD corresponds to the encoding mode data EN1 / EN2 corresponding to the first encoding mode or the second encoding mode from the encoding mode determining circuit 34. It is made.

As described above, the encoding processing mode determination circuit 34 uses the image data in macroblock units read from the frame memory group 10 to determine the first encoding processing mode and the second encoding processing mode. It is determined which of the encoding processing modes is more efficient in encoding, and the encoding mode data EN1 or EN2 according to the determination result is determined.
Is output. More specifically, the encoding processing mode determination circuit 34 calculates, for example, the sum of absolute value differences of each pixel between an odd field (first field) and an even field (second field) of each frame. If the value of the sum of absolute differences is, for example, less than a certain threshold value T0 (that is, when the motion of the image is small), a code indicating that the encoding in the first encoding mode is more efficient. If the sum of the absolute value difference is equal to or greater than the threshold value T0 (when the motion of the image is large), the encoding in the second encoding mode is performed. Encoding mode data EN2 indicating that the
Is output.

Note that the above-mentioned encoding processing mode determination circuit 34
At the time of the determination in the field motion detection circuit 2
It is also possible to make a determination using the motion vector data FDMV from 1. That is, if the motion vector data FDMV between the odd field and the even field is smaller than a certain threshold value t0, the first encoding processing mode is selected,
Conversely, if it is equal to or greater than the threshold value t0, it is possible to select the second encoding processing mode.

The encoding mode data EN1 / EN2 from the encoding processing mode judging circuit 34 is sent to the frame / field mode judging circuit 33,
From the frame / field mode determination circuit 33,
The frame orthogonal transform mode data MDFM or the field orthogonal transform mode data MDFD corresponding to the encoding mode data EN1 / EN2 is output.

That is, when the encoding mode data from the encoding processing mode determination circuit 34 is the data EN1 indicating the first encoding processing mode, the frame / field mode determination circuit 33 A process of adaptively switching the frame orthogonal transform mode or the field orthogonal transform mode is performed for each macroblock. Accordingly, the frame / field mode determination circuit 33 outputs the adaptively switched frame orthogonal transform mode MDFM or field orthogonal transform mode data MDFD.

On the other hand, if the encoding mode data from the encoding mode determining circuit 34 is the data EN2 indicating the second encoding mode, the frame / field mode determining circuit 33 As described above, the blocking of the orthogonal transform of all the macroblocks in one frame is performed in the field orthogonal transform mode. Accordingly, the field orthogonal transform mode data MDFD is output from the frame / field mode determination circuit 33.

The mode data MDFM / MDFD for blocking the orthogonal transform of either the frame / field output from the frame / field mode determination circuit 33, and
The encoding mode data EN1 / EN2 from the encoding processing mode determination circuit 34 is supplied to the address generator 35.
And transmitted to the frame memory group 20 with motion compensator.
The orthogonal transform mode data (MDFM / MDFD), the motion prediction mode data (MPFM / MPFD), the encoding mode data EN1 / EN2, and the motion vector data (FMMV / FDMV) are the same as those of the above-described variable length code. It is also sent to the conversion circuit 15.

The address generator 35 outputs the orthogonal transform mode data MDFM / MDFD for each macroblock.
The frame memory group 10 is controlled so as to output image data of a macroblock divided into blocks according to the encoding mode data EN1 / EN2. That is, as described above, for example, when the encoding mode data EN1 / EN2 is the data EN1 indicating the first encoding processing mode, the address generator 35 blocks the orthogonal transform in the odd cycle. Mode (data MD
FM / MDFD) to control the frame memory group 10 so as to output a macroblock that is divided into blocks. The frame memory group 10 is controlled so as to output a macroblock which is divided into blocks corresponding to the field orthogonal transform mode (data MDFD).

In other words, for example, when the first encoding processing mode is selected and the encoding mode data EN1 is supplied to the address generator 35, for example, the orthogonal transformation mode data is stored in frame units. DC
If the data MDFM indicates the T processing, the address generator 35 operates as shown in FIG. The block memory group 10 is controlled so as to output (block). That is, in this case, the address generator 34
Represents a macroblock consisting of 1 to 16 lines as shown in FIG.
The frame memory group 10 is controlled so that the data is divided into 16 lines and four 8 × 8 blocks (macroblocks) are output.

When the first encoding processing mode is selected and the encoding mode data EN1 is supplied to the address generator 35, for example, the orthogonal transform mode data indicates DCT processing in field units. If the data is MDFD, the address generator 35 scans the even and odd scans separately as shown in FIG. 5, and scans the macroblocks (macroblocks in odd and even fields in field units). Is controlled to output the frame memory group 10. That is, the address generator 34 is provided in the above-described FIG.
As shown in FIG. 1, 1 line, 3 lines, 5 lines, 7 lines, 9 lines, 11 lines, 13 lines, 15 lines (odd field or each line of the first field),
It is divided into two lines, four lines, six lines, eight lines, ten lines, twelve lines, fourteen lines, and sixteen lines (even lines or each line of the second field), and these odd fields and even fields respectively. The frame memory group 10 is controlled so that two 8 × 8 blocks (macro blocks) are output.

In FIGS. 4 and 5, odd lines are indicated by solid lines, and even lines are indicated by dotted lines.

Further, for example, when the second encoding processing mode is selected and the encoding mode data EN2 is supplied to the address generator 35, the address generator 35 is connected to the odd number as described above. The frame memory group 10 is controlled so as to output a macroblock that is divided into blocks in the cycle and the even cycle in accordance with the field orthogonal transform mode. That is, the address generator 3 when the second encoding mode is selected.
Reference numeral 5 controls the frame memory group 10 so that two 8 × 8 blocks are always output (however, only a luminance component will be described later). More specifically, in the odd cycle, the address generator 35 includes two 8 × 8 blocks of macroblocks in one odd field only.
The frame memory group 10 is controlled so as to output the frames (for one screen). Then, in the even cycle, only the even fields have two 8 × 8 macroblocks for one frame (for one screen). ) The frame memory group 10 is controlled so as to be output.

As described above, the image data output from the frame memory group 10 controlled by the address generator 35 is subjected to the DCT processing by the DCT circuit 13 as described above. That is, for example, when the first encoding processing mode is selected and the frame orthogonal transform mode is selected, the DCT circuit 13 performs the DCT transform on the 8 × 8 pixel unit block as shown in FIG. I do. Also,
For example, when the first encoding processing mode is selected and the field orthogonal transform mode is selected, the DCT circuit 13 performs the DCT transform on the 8 × 8 pixel unit block as shown in FIG. When the second encoding processing mode is selected, as described above, only the odd field is used in the 8 × 8 block in the odd cycle in the odd cycle.
The CT transform is performed, and the DCT transform is performed in the 8 × 8 block only in the even field in the even cycle.

Further, the motion prediction mode data MPFM / MP from the frame / field mode determination circuit 33
FD and orthogonal transformation mode data MDFM / MDFD and the motion vector data FMMV / F selected by the selector 24
The DMV and the encoding mode data EN1 / EN2 from the encoding processing mode determination circuit 34 are also supplied to the frame memory group with motion compensator 20. Therefore, the frame memory group with motion compensator 2
0, the motion prediction mode data M
Orthogonal transform mode data MDFM / MDFD and encoding mode data EN1 / E in PFM / MPFD and DCT processing
N2 and the motion vector data FMMV /
Motion compensation using FDMV is performed.

As described above, in the first encoding processing mode and in the frame orthogonal transform mode, for example, the motion prediction of the P and B frames is performed as shown in FIG. Motion is predicted from the frame. Therefore, in the DCT circuit 13, the difference image from the predicted image is subjected to DCT (DCT conversion in a unit block of 8 × 8 pixels). FIG. 6 shows the previous frame,
The current frame and the subsequent frame are shown, the arrow in the figure indicates a motion vector, and the MB indicates a macroblock. In the motion prediction of the P and B frames in the first encoding processing mode and in the field orthogonal transform mode, FIG.
As shown in (1), for each of the odd field and the even field of the macroblock, motion prediction is performed from the odd or even field (first or second field) of the previous and subsequent frames. Therefore, in the DCT circuit 13, the difference image from the predicted image is subjected to DCT (DCT conversion in a unit block of 8 × 8 pixels). FIG. 7 shows the odd field and the even field of the previous frame, the current frame, and the subsequent frame, respectively.
In the figure, arrows indicate motion vectors, and MB indicates macroblocks.

Further, the motion prediction in the field orthogonal transform mode in the second encoding mode is as follows:
For example, as shown in FIG. 8, for each of the odd field and the even field of the macroblock, the motion is predicted from the odd or even field of the previous and subsequent frames, and the motion between the fields in each frame is also predicted. . Therefore, in the DCT circuit 13, the difference image from the predicted image is subjected to DCT (DCT conversion in a unit block of 8 × 8 pixels). FIG. 8 illustrates the odd field and the even field of the previous frame, the current frame, and the subsequent frame. Arrows in the figure indicate motion vectors, and MB indicates a macroblock.

As described above, the high-efficiency image signal encoding apparatus according to the first embodiment has the following features in accordance with the first and second encoding processing modes (that is, the magnitude of image motion). Since the coding is switched between the coding that does not perform the inter-field prediction in the frame and the coding that performs the inter-field prediction in the frame, the most efficient coding is possible. In particular, the second encoding processing mode is effective for frames with large motion.

By the way, the encoding apparatus of the first embodiment specifically performs the following motion prediction and DCT transform processing for each digital VTR format.

Here, in FIGS. 10, 11 and 12, the fields constituting the frame of the I frame (I picture) are represented by Io field (odd field of I frame) and Ie field (even field of I frame). And the fields constituting the P frame (P picture) are Po field (odd field) and P field.
An e field (even field) is used, and fields constituting the B frame (B picture) are a Bo field (odd field) and a Be field (even field).

In this embodiment, the frame orthogonal transformation mode of the orthogonal transformation into blocks is constructed by combining the odd fields and the even fields as shown in FIG. This is a mode in which the macroblock is configured) and this macroblock is used as a processing unit. In addition, the field orthogonal transformation mode of the orthogonal transformation into blocks is divided into an odd field and an even field as shown in FIG. In this mode, a macro block is configured (that is, a macro block is configured for each field), and this macro block is used as a processing unit. For this reason, for example, in the I frame, the frame orthogonal transform mode and the field orthogonal transform mode are switched for each macroblock.

Further, in this embodiment, for one frame, the above-mentioned odd cycle (odd cycle) of the period during which the encoding process scans the odd field in the interlace, and the above even number of the period during which the scanning of the even field is performed. Cycle (even cycle).

Therefore, in the case of the present embodiment, for example, a digital VT of a so-called 4: 2: 0 component
When the R format is used, for example, as shown in FIG. 9, when the above-described orthogonal transformation is in the frame orthogonal transformation mode, for example, luminance blocks Y0, Y1, Y2, and Y3 composed of odd and even fields are used. DCT processing is performed on each unit block of the macro block including the color difference blocks Cb0 and Cr1 of the odd field. On the other hand, when the orthogonal transform is blocked in the field orthogonal transform mode, the luminance blocks Y02o and Y13o of each odd-numbered field and the luminance block Y02 of each even-numbered field are used.
e, Y13e and the color difference block Cb of the odd field.
D of each unit block of the macroblock consisting of 0 and Cr1
CT processing is performed.

In the first encoding processing mode of this embodiment, when the frame motion prediction mode is set,
As shown in (1), for example, a motion prediction MCP between an I frame and a P frame can be performed. On the other hand, in the field motion prediction mode, the motion prediction MCoPo between the Io field and the Po field, the motion prediction MCoPe between the Io field and the Pe field, and the Ie field and Po
A motion prediction MCe Po between the fields and a motion prediction MCe Pe between the Ie and Pe fields are possible. That is, in the case of FIG. 10, the motion prediction and orthogonal transformation modes can exist independently in the frame motion prediction mode and the frame orthogonal transformation mode, and in the field motion prediction mode and the field orthogonal transformation mode, respectively. One vector is obtained, and two motion vectors are obtained in the field motion prediction mode.

Therefore, in the first encoding processing mode of the first embodiment, for example, when the blocking of the orthogonal transform of the I frame is the frame orthogonal transform mode, the odd cycle (odd cycle) is used. , The Io field and the Ie field are combined to form the macroblock. For example, in the odd cycle, DCT conversion is performed for each macroblock (DCT is 8 × 8).
Is performed for each unit block described above), quantization, and variable length coding are performed. On the other hand, no data is sent in the even cycle of the frame orthogonal transform mode.

In the first encoding mode of the present embodiment, when the orthogonal transform is blocked in the field orthogonal transform mode, the Io field and the Ie field are separately separated in the odd cycle. The above-mentioned macro block is configured, and the DC
T transform (DCT is performed for each of the 8 × 8 unit blocks), quantization, and variable length coding are performed. On the other hand, in the even cycle of the field orthogonal transform mode, no data is transmitted as can be seen from FIG.

From the above, as shown in FIG. 11, in the case of the P frame in the first encoding processing mode, the following processing is performed. For example, when the orthogonal transform of a P frame is blocked in the above-described frame orthogonal transform mode and the motion prediction is in the above-described frame motion predictive mode, in the odd cycle, the reference image is used as a forward image (I-frame image). , And the difference from the original image is encoded using the macroblock in which the Io field and the Ie field are alternately combined as a predicted image. On the other hand, no data is sent in the even cycle in these modes.

In the first encoding processing mode, when the orthogonal transform of the P frame is blocked in the frame orthogonal transform mode and the motion prediction is the field motion prediction mode, the Io field and the Ie field are used in the odd cycle. Using the fields (or Po and Pe fields) as reference images, respectively, a motion vector MVo Po between the Io field and the Po field, and a motion vector MVe Po, Io between the Ie field and the Po field.
Motion vector MV between field and Pe field
o Find the motion vector MVePe between the Pe, Ie and Pe fields, and predict both the odd and even field predictions (eg, the average of the even and odd field predictions). The prediction that minimizes the prediction error from the current P frame is selected.
The macroblock in which the o field and the Ie field are combined is used as a prediction image to encode the difference from the original image. On the other hand, no data is sent in the even cycle of this mode.

Further, in the first encoding mode, P
When the orthogonal transform of a frame is blocked in the field orthogonal transform mode and the motion prediction is the frame motion predictive mode, in the odd cycle, the reference image is an I-frame image (or a P-frame image) and A motion vector MVP is detected, and the difference between the macroblock composed of the Io field and the Ie field separately divided and the original image (macroblock composed of the Po field and the Pe field divided separately) is determined as a predicted image. Encode. On the other hand, in the even cycle of this mode, no data is transmitted as described above.

In the first encoding processing mode, when the orthogonal transform of the P frame is blocked in the field orthogonal transform mode and the motion prediction is in the field motion prediction mode, the Io field and Ie are used in the odd cycle.
Using the fields (or Po and Pe fields) as reference images, respectively, a motion vector MVo Po between the Io field and the Po field, and a motion vector MVe Po and IV between the Ie field and the Po field.
the motion vector M between the o field and the Pe field
Vo Pe, the motion vector MVe Pe between the Ie and Pe fields is detected, and the prediction of the odd field and the prediction of the even field is performed (for example, the average of the prediction of the even field and the prediction of the odd field). ,
Select the prediction that minimizes the prediction error with the current P frame,
The macroblock composed of the Io field and the Ie field separately divided is used as a predicted image as an original image (P
The difference between the o field and the macro block composed of the Pe field separated separately is encoded. On the other hand, no data is sent in the even cycle of this mode.

Further, in the case of the B frame in the first encoding processing mode, the following processing is performed. For example, when the blocking of the orthogonal transformation of the B frame is the frame orthogonal transformation mode and the motion prediction is the frame motion prediction mode, in the odd cycle, the reference image is a front and rear image, and a motion vector between frames, that is, , I
Motion vector FMVB between a frame and a B frame and motion vector BMV between a P frame and a B frame
B, and among the forward prediction, backward prediction, and bidirectional prediction (average of forward prediction and backward prediction), a prediction that minimizes the prediction error with respect to the current frame is selected, and the odd field and the even field alternate. The difference from the original image is encoded using the combined macroblock as a prediction image. On the other hand, no data is sent in the even cycle of this mode.

In the first encoding processing mode, when the orthogonal transformation of the B frame is blocked in the frame orthogonal transformation mode and the motion prediction is the field motion prediction mode, the reference image is forwarded in the odd cycle. The prediction of the odd field and the prediction of the even field are performed on these images as the backward and forward images, respectively, and the respective motion vectors, ie, the motion vectors FMVo Bo, Ie and Bo fields between the Io and Bo fields, Between the motion vectors FMVeBo, Io
Motion vector F between the field and the Be field
MVoBe, motion vector FMVeBe between the Ie and Be fields, motion vector BMVoBo between the Po and Bo fields, motion vector BMVeB between the Pe and Bo fields.
o, a motion vector BMVoBe between the Po field and the Be field, and a motion vector BMVeBe between the Pe field and the Be field, and the prediction of the odd field and the prediction of the even field by the respective vectors are performed. (For example, the average of the prediction of the even field and the prediction of the odd field), the prediction that minimizes the prediction error from the current frame is selected, and the Io field and the Io field are selected.
e field (or Po and Pe fields)
And the difference from the original image is encoded using the macroblock combined with On the other hand, no data is sent in the even cycle of this mode.

Further, in the first encoding mode, B
When the orthogonal transform of a frame is blocked in the field orthogonal transform mode and the motion prediction is the frame motion predictive mode, in the odd cycle, a reference image is used as a forward and backward image, and a motion vector between frames, that is, an I frame Motion vectors FMVB and P between frame and B frame
A motion vector BMVB between a frame and a B frame is detected, and a prediction that minimizes a prediction error with respect to a current frame is selected from forward prediction, backward prediction, and bidirectional prediction (average of forward prediction and backward prediction). Then, the difference from the original image is encoded using the above-mentioned macroblock, which is configured by separately dividing the odd field and the even field, as a prediction image. On the other hand, no data is sent in the even cycle of this mode.

In the first encoding processing mode, when the orthogonal transform of the B frame is blocked in the field orthogonal transform mode and the motion prediction is the field motion prediction mode, the reference image is set to the front in the odd cycle. For these images as the subsequent images, prediction of the odd field and prediction of the even field are performed, and the respective motion vectors, that is, the motion vectors FMVo Bo, the Ie field and the Bo field between the Io field and the Bo field are calculated. Between the motion vectors FMVeBo, Io
Motion vector F between the field and the Be field
MVoBe, motion vector FMVeBe between the Ie and Be fields, motion vector BMVoBo between the Po and Bo fields, motion vector BMVeB between the Pe and Bo fields.
o, a motion vector BMVoBe between the Po field and the Be field, and a motion vector BMVeBe between the Pe field and the Be field, and the prediction of the odd field and the prediction of the even field by the respective vectors are performed. (For example, the average of the prediction of the even field and the prediction of the odd field), the prediction that minimizes the prediction error from the current frame is selected, and the Io field and the Io field are selected.
e field (or Po and Pe fields)
Encodes the difference from the original image using the macroblock separately configured as a predicted image. On the other hand, no data is sent in the even cycle of this mode.

However, in the case of the first encoding processing mode of this embodiment, as can be seen from FIG. 10, the motion prediction between the Io field and the Ie field and the motion prediction between the Po field and the Pe field are performed. The prediction and the motion prediction between the Bo field and the Be field cannot be performed.

In this case, by using the second encoding processing mode of the present embodiment, prediction from an odd field to an even field can be performed in each picture. That is, in the field orthogonal transform mode in the second encoding processing mode, each unit of each luminance block Y02o, Y13o of the odd field and each color difference block Cb0, Cr1 of the odd field in the odd cycle. DCT processing the block. Thereafter, each unit block of each of the luminance blocks Y02e and Y13e in the even field is subjected to DCT processing in an even cycle.

As shown in FIG. 12, the motion prediction in the case of the second encoding processing mode is as shown in FIG.
In addition to Pe, motion prediction SMCI between the Io field and the Io field, and motion prediction SMCP between the Po field and the Pe field are possible.

Accordingly, in the case of the field orthogonal transform mode of the second encoding processing mode, only the odd field of the macroblock is encoded in the odd cycle, and only the even field of the macroblock is encoded in the even cycle. . Thus, for example, at the end of the odd cycle, the decoder side described later
o The whole field will be obtained. Therefore, in the even cycle of the I frame, as shown in FIG. 11, motion prediction is performed using the Io field as a reference image for the macroblock of the Ie field in the field orthogonal transform mode, and the motion vector SMVI and the prediction image are used. Can be encoded.

Further, as shown in FIG. 11, in the case of a P frame in the second encoding processing mode, the following processing is performed. For example, when the motion prediction of the P frame is in the frame motion prediction mode, a motion vector MVP between frames is detected after the odd cycle and the even cycle by using the reference image as a forward image (I-frame image). The difference from the original image is encoded using the macroblock in which the field and the Ie field are combined as a predicted image.

When the motion prediction of the P frame in the second encoding processing mode is the field motion prediction mode, in the above-mentioned odd cycle, the Io field and the Ie field are used as reference image images, respectively. Motion vector MVOPo, Ie field and P
The motion vector MVe Po between the o-field and the o-field is detected. In the even cycle of this mode, a motion vector MVoPe between the Io field and the Pe field, a motion vector MVe Pe between the Ie field and the Pe field, and a motion vector MPoPe and the Po field and the Pe for the macroblock in the field orthogonal transform mode. A motion vector SMVP between the respective fields is detected, and prediction is performed from among prediction of an odd field, prediction of an even field, prediction of an odd field of the current frame, and prediction of an average of two predictions among them. The prediction that minimizes the error is selected, and the difference from the predicted image is encoded.

Further, for example, when the motion prediction of the B frame is the frame motion prediction mode in the second encoding processing mode, after the odd cycle and the even cycle, the reference image is set to the front and rear images, and the motion between the frames is determined. Vector, that is, the motion vector FMVB between the I frame and the B frame and the motion vector BMVB between the P frame and the B frame are detected, and forward prediction, backward prediction, bidirectional prediction (average of forward prediction and backward prediction), and Of these, the prediction that minimizes the prediction error from the current frame is selected, and the difference from the predicted image is encoded.

In the second encoding processing mode, when the motion prediction of the B frame is the field motion prediction mode,
In the above-mentioned odd cycle, prediction of odd fields and prediction of even fields are performed on these images, with the reference images being forward and backward, respectively. Motion vector FMVe Bo between the field, Po field and Bo
Motion vector BMV between fields Bo, Bo, Pe Motion vector between fields BMV
eBo is detected. Thereafter, in the same manner as described above, the prediction that minimizes the prediction error is selected, and the difference from the prediction image is encoded. Further, in the even cycle of this mode, Io
Motion vector FM between field and Be field
VoBe, a motion vector FMVeBe between the Ie and Be fields, a motion vector BMVoBe between the Po and Be fields, and a motion vector BMVeBe between the Pe and Be fields.
, And the prediction of the odd field of the current frame (that is, the prediction by the motion vector SMVB between the Bo field and the Be field) is also performed, and the prediction with the minimum prediction error is selected. Encode the difference.

Further, in the case of this embodiment, for example,
When handling the so-called 4: 2: 2 component digital VTR format, for example, as shown in FIG. 13, in the frame orthogonal transform mode, for example, luminance blocks Y0 and Y1 composed of odd and even fields are used.
, Y2, Y3, and the DCT processing of each unit block of a macroblock composed of color difference blocks Cb01, Cr01, Cb23, and Cr23 composed of odd and even fields. In the field orthogonal transform mode, each of the luminance fields Y02o and Y13o of the odd field and the color difference blocks Cb0123o and Cr0123o of the odd field, the luminance blocks Y02e and Y13e of the even field, and the color difference blocks Cb0123e and Cr0123e of the even field. DCT processing is performed on each unit block of the macro block.

The motion prediction in the first encoding mode in the case of the example of FIG. 13 is as shown in FIG. 10 described above. However, in this case as well, the motion prediction between the Io field and the Ie field, the motion prediction between the Po field and the Pe field, and the motion prediction between the Bo field and the Be field cannot be performed, as described above.

Therefore, in this case, as described above, the second encoding mode may be used. That is, in the field orthogonal transform mode in the second encoding processing mode, each unit of the luminance blocks Y02o, Y13o of the odd field and the color difference blocks Cb0123o, Cr0123o of the odd field in the odd cycle. DCT processing the block. Thereafter, each unit block of the luminance blocks Y02e and Y13e of the even field and the chrominance blocks Cb0123e and Cr0123e of the even field is subjected to DCT processing in an even cycle.

The motion prediction in the case of the example of FIG.
Same as 1.

Further, in the present embodiment, the above 4: 2: 2
When handling the digital VTR format of the component, for example, in addition to the processing as shown in FIG.
As shown in FIG. 14, the motion prediction of a frame is performed in units of macroblocks MB. When the motion prediction of a field is performed, a certain macroblock MB (i, j) and a macroblock MB ( i + 1, j) as a set, and it is also possible to perform the motion prediction of the odd field and the motion of the even field with respect to the set MBg of the macroblock.

FIG. 15 shows a part of the macroblock extracted from the frame in the case of the example shown in FIG. It is assumed that the process proceeds in the direction of the arrow in FIG. That is, FIG. 15 shows that a certain macroblock MB (i, j) has the next macroblock MB (i, j + 1) and the macroblock MB (of the next line) located below them (of the next line). i + 1,
j) and MB (i + 1, j + 1).

In the macroblock shown in FIG. 15, for example, in the case of the frame orthogonal transformation mode, each macroblock MB (i, j), MB (i, j + 1),.
B (i + 1, j), MB (i + 1, j + 1)... Each luminance block Y0, Y1 and color difference blocks Cb01, Cr01 are DCT
It is processed. Therefore, in the case of the frame orthogonal transform mode, the processing of each macroblock is not affected by the orthogonal transform mode of another macroblock.

On the other hand, in the case of the field orthogonal transform mode, as shown in FIG. 16, the macroblocks constituting the macroblock set MBg are replaced with the macroblocks MBgo of the odd-numbered field with respect to the macroblock set MBg. And the even-numbered field macroblock MBge, and the luminance blocks Y0o, Y1o and the color difference blocks Cb01o, Cr0 in the odd-numbered field macroblock MBgo.
DCO processing is performed on 1o. Here, for example, the set MBg of the macroblock corresponds to the macroblock MB (i,
j) and MB (i + 1, j), the luminance blocks Y0o and Y1o in the macroblock MBgo of the odd-numbered field in the macroblock MBg correspond to the macroblock MB (i, j). ), And the luminance block of the odd field of the macroblock MB (i + 1, j), and the chrominance blocks Cb01o, Cb01o, in the macroblock MBgo of the odd field.
Cr01o is a color difference block of an odd field of the macro block MB (i, j) and the macro block MB (i, j).
+1) and a color difference block of an odd field of (j). Similarly, the luminance blocks Y0e and Y1e in the even-field macroblock MBge are the even-field luminance blocks of the macroblock MB (i, j) and the even blocks of the macroblock MB (i + 1, j). Chrominance blocks Cb01e and Cb01e in the macroblock MBge of the even field
r01e is composed of a color difference block of the even field of the macroblock MB (i, j) and a color difference block of the even field of the macroblock MB (i + 1, j).

As described above, motion estimation and DC
The relationship with each mode of T conversion is as described below. That is, in the encoding apparatus according to the present embodiment, for example, when the macroblock MB (i, j) is subjected to motion prediction in the frame orthogonal transform mode and is subjected to DCT transform in the frame orthogonal transform mode, for example, the motion compensator An image decoded in the attached frame memory group 20 is used as a reference frame, and a predicted image extracted from the reference frame is
The difference from the input image (original image) is DCT-transformed. Then, the DCT coefficient and the frame motion vector are transmitted.

Also, for example, the macro block MB
In (i, j), when the motion prediction is in the field motion prediction mode and the DCT transform is in the field orthogonal transform mode, in the macroblock MB (i, j), the prediction image extracted from the odd field and the original of the odd field are used. The difference from the image and the motion vector of the odd field are encoded. In the macroblock MB (i + 1, j), the difference between the predicted image extracted from the even field and the original image of the even field, and the motion vector of the even field are encoded.

Further, for example, the macro block MB
In (i, j), when the motion prediction is the field motion prediction mode and the DCT transform is the frame orthogonal transform mode,
In the macroblock MB (i, j), the frame difference between the predicted image and the input image with respect to the position of the macroblock MB (i, j) extracted from the reference frame, the motion vector of the odd field and the motion vector of the even field Is transmitted. Also, in the macro block MB (i + 1, j), the macro block M extracted from the reference frame
The frame difference between the predicted image and the input image for the position of B (i + 1, j) is transmitted.

Further, for example, the macro block M
In B (i, j), when the motion prediction is in the frame motion prediction mode and the DCT transform is in the field orthogonal transform mode, in the macroblock MB (i, j), the predicted image extracted from the odd field and the The difference from the original image, the frame motion vector of the macroblock MB (i, j), and the frame motion vector of the macroblock MB (i + 1, j) are transmitted. In the macroblock MB (i + 1, j), the difference between the predicted image of the odd field and the input image is transmitted.

In the coding apparatus of the present embodiment, the present code is realized by adding an extension bit to the conventional macroblock type to obtain compatibility with the conventional macroblock type.

That is, in the case of the first embodiment, for example, in a B frame, there are three macroblock types, ie, pre-prediction, post-prediction, and both predictions as described above. Since there are two types of prediction from the odd field and the even field of the previous frame, the present code is realized by adding an extension bit for recognizing either prediction. The prediction in this case is 2
Therefore, the extension bit may be added by one bit in one direction (previous and post-prediction). For example, in the case of prediction from an odd field in the pre- or post-prediction, the code “1”
In the case of prediction from an even field, the code "0" may be added as an extension bit to the conventional macroblock type. Also, in both predictions, both extension bits are added for the pre- or post-prediction.

In the frame motion prediction mode, the conventional bit stream (MPE)
It has the same format as G).

The above is similarly applied to the case of a P frame.

Next, in the case of the present embodiment, for example, in the B frame, the macroblock type includes pre-prediction, post-prediction, and both predictions as described above. Predictions from
An extension bit must be added to the macroblock type to recognize whether prediction is from an even field or from an odd field in its own frame. That is, in the field motion prediction mode of the previous prediction, since there is a prediction from within the own frame, one or two extended bits are necessary to express three types of predictions including the odd and even numbers with the extended bits. In the post-prediction field motion prediction mode, there are only two types, odd and even, so that one extension bit is always required. For example, in the previous prediction,
A code “1” is added for prediction from an odd field of the previous frame, a code “01” is added for prediction from an even field of the previous frame, and a code “11” is added for prediction from an odd field of the current frame. , In the post-prediction, the code “1” is used in the case of prediction from an odd field of the subsequent frame,
In the case of prediction from the even field of the subsequent frame, the code "0" may be added to the conventional macroblock type as an extension bit.

In the mode of processing a frame, no extension bits are added, and the format is the same as that of a conventional bit stream (MPEG). Also, in both predictions, both extension bits are added for the pre- or post-prediction.

The above is similarly applied to the case of a P frame.

Further, as a modification, the number of extension bits in the case of the pre-prediction can be reduced to one bit. That is, in the even cycle in the field motion prediction mode, as shown in FIG. 17, the prediction indicated by the dashed line in the figure from the odd field of the previous frame which is farthest in time and position is abolished. Can be reduced to two, and the previous motion prediction mode can be transmitted with 1-bit extension. Specifically, in the previous prediction in the odd cycle, the code “1” is used in the case of prediction from an odd field of the previous frame,
In the case of prediction from the even field of the previous frame, the code is “0”. In the prediction in the even cycle, the code is “1” in the case of prediction from the odd field of the current frame. In the case of prediction from the odd field of the subsequent frame, the code "0" is used as the extension bit, and in the case of the prediction from the even field of the subsequent frame, the code "0" is used as the extension bit. May be added to the macro block type.

FIG. 18 shows a configuration example of the encoding apparatus according to the second embodiment. Note that, in FIG. 18, the same components as those in FIG. 1 described above are denoted by the same reference numerals, and a detailed description thereof will be omitted.

The configuration of the apparatus of the second embodiment is a three-pass encoding apparatus, which performs processing three times to process one frame.

That is, in the second embodiment, the first pass performs the processing of the second encoding processing mode (with intra-frame prediction between frames) using the fixed quantization width, and the second pass performs the fixed pass. The processing of the first encoding processing mode (without intra-frame inter-frame prediction) based on the quantization width is performed, and in the third pass, the processing in which the number of generated bits is small in the first and second passes is selected. Processing is performed by controlling the quantization width.

In the second embodiment, a macroblock converter 55, a changeover switch 57, a field block conversion circuit 56, and a changeover switch 58, which will be described later, are inserted and connected to the frame memory group 10 at a later stage. You.
The image data from the frame memory group 10 is sent to a motion detection circuit 51 that performs frame and field motion detection. An output from the motion detection circuit 51 is sent to a mode determination circuit 52 for selecting a frame / field mode for blocking and motion prediction of orthogonal transform, the frame memory group 20 and the variable length coding circuit 15.

The output mode data from the mode determination circuit 52 is sent to the frame memory group 20 and the variable length coding circuit 15, and the field orthogonal transform mode data is sent to one input terminal of a two-input AND gate 53. Can be The other input terminal of the two-input AND gate 53 has a changeover switch 59 that is switched via the inverter 54 in accordance with the first, second, and third passes.
Output is supplied. The output terminal of the 2-input AND gate is connected to the changeover switch 57,
58 switching control terminals.

The data of the number of generated bits is output from the variable length coding circuit 15, and based on the data of the number of generated bits, any one of the first and second encoding modes has a smaller number of generated bits. Is sent to a selection circuit (intra-frame inter-field prediction presence / absence determination circuit) 60 for selecting one of the modes. Further, the accumulated amount data from the buffer 16 is transmitted to the changeover switch 6 together with the variable length encoding circuit 15.
1 is supplied to one of the switched terminals. A fixed value of the first and second passes is supplied to the other switched terminal of the changeover switch 61.

In the apparatus of the second embodiment, the image input to the terminal 1 is once stored in the frame memory group 10, and the necessary frame or field data is called out from the frame memory group 10. A motion vector is obtained by the motion detector 51 using these image data. The mode determination circuit 52 determines the mode of the field / frame for each macroblock from the motion prediction residual from the motion detector 51. The macroblock converter 55 connected to the subsequent stage of the frame memory group 10 outputs information corresponding to the first, second, and third passes through the changeover switch 59 (that is, the first encoding process). Mode or the information on the presence / absence of intra-frame inter-frame prediction, which is the second encoding processing mode), and when information on the second encoding processing mode is received as this information, the odd field (the first field) ) Is transmitted, and then the even field (second field) is transmitted, and the blocking in the frame orthogonal transform mode is turned off. Further, the image data in which the macroblock is determined to be a block in the frame orthogonal transformation mode based on the information of the first encoding processing mode in the macroblock generator 55 is determined based on the residual from the motion detector 51. If the circuit 52 determines that the mode is the field mode, the field block conversion circuit 56 converts the block into a block in the frame orthogonal transform mode.

In the first pass and the second pass, encoding is performed with a fixed quantization width, and the bit generation amount is compared by the selection circuit 60. Is selected for each frame, and actual encoding is performed in the third pass. At this time, the information of the selected mode is added by one bit for each frame.

FIG. 19 is a block diagram of a picture signal decoder. That is, the high-efficiency decoding apparatus of the present embodiment
Upon receiving header information including encoded image data to be reproduced, detected motion vector information, motion prediction mode information, blocking mode information (orthogonal transformation mode information), and encoding processing mode information (encoding processing mode data), Decrypt and
An inverse variable length encoding circuit 51 that outputs the detected motion vector information, the motion prediction mode information, the orthogonal transform mode information, and the encoding processing mode data of the header information together with the decoded image decoded data; From the processing mode data, the address increment value in the frame buffers 61, 62, 64 for storing the image decoding data is calculated, and the head address of each macro block is obtained.
The start address is stored in the frame buffers 61, 62, 6
4 and the relative addresses of the macroblocks other than the head address are added to the frame buffers 61, 62, and 64 to access data, and the detected motion vector and the motion prediction mode are added. Information, the orthogonal transformation mode information, and the encoding processing mode data, execute a motion compensation frame or inter-field prediction corresponding to the mode information, and transfer the motion-compensated image signal to the frame buffers 61 and 6.
Motion compensation circuits 59, 6 configured to send the
0, 63, 65, and 66.

In FIG. 20, data encoded by the high-efficiency encoding apparatus of the above embodiment is temporarily recorded on a storage medium such as a CD. The coded data reproduced from the CD or the like is first decoded by the inverse variable length coding circuit 51 via the input terminal 50 into header information and the like for each sequence, each frame group, and each frame. In the odd cycle of the frame, header information is decoded for each slice (group of macroblocks), and the quantization width is included in the header of this slice. Then, for each macroblock, the macroblock address, frame / field motion prediction mode and frame / field orthogonal transform mode information, coding mode data, and macroblock type indicating the decoding method are decoded, and the quantization width is Decrypted when updated.

If the macroblock is in the frame orthogonal transform mode, the entire macroblock is decoded in the odd cycle and nothing is decoded in the even cycle. When the blocking is performed in the field orthogonal transform mode, only the block including the odd field in the macroblock is decoded in the odd cycle, and the block including the even field is decoded in the even cycle.

The image information is decoded via an inverse quantizer 53 for performing an inverse quantization process and an inverse DCT circuit 54 for performing an inverse DCT transform process, and it is determined whether or not the image is a difference image according to a macroblock type. According to the determination result, the adder 56 switches the mode switch 57 for switching between addition and non-addition to the reference image (corresponding to non-intra / intra in MPEG encoding). The decoded image is input to the frame buffer 64 or 61 in the case of an I frame or a P frame (alternately every time the I frame and the P frame are processed), and
In the case of a frame, it is input to the frame buffer 62.
Each frame buffer is composed of two field buffers, and odd / even field images are stored separately in the respective field buffers. The writing to the frame buffer is controlled by switching the switch 58.

At this time, the frame buffers 61, 62,
The addresses written to 64 are address generators 81, 8
2,83. This address generator 81,
At 82 and 83, the frame buffers 61, 62, and 62 are extracted from the encoding processing mode data in the header information of the macroblock.
The address increment value at 64 is calculated, and the start address of each macro block is obtained.

Further, the quantization width data is stored in the memory 52 for one field. This quantization width data is sent to the inverse quantizer 53 via the switch 55 that is switched according to the output of the inverse variable length encoding circuit 51. Here, in the even cycle, since only the macroblock processed in the field orthogonal transform mode is decoded, the macroblock type of the macroblock address to be decoded for each macroblock and the motion vector required for the prediction method indicated by the macroblock type are: The decoded and further transmitted difference image is added to the motion-compensated image from the reference field to obtain a reproduced image.

Each of the frame buffers 64, 6
The data of the motion compensation processing circuits 65, 66,
The motion is compensated by 59, 60 and 63. At this time, each motion compensation circuit switches between frame motion compensation and field motion compensation according to the orthogonal transform mode (frame / field) in DCT processing.

These motion-compensated images are sent to the selected terminals of the changeover selection switches 67, 68, 71. These changeover selection switches 67, 68, 71 are switched so that a reference field or a frame indicated by a macroblock type decoding method can be extracted. Here, the switch selection switch 71 is supplied with a signal which is obtained by adding the outputs of the switch selection switches 67 and 68 by the adder 69 and then halving the result by the divider 70, and the output of the switch 67. Is done. The output of the switch 71 is sent to the switch 57.

Further, each of the frame buffers 64, 61, 6
The output of 2 is sent to the display 73 via the changeover selection switch 72. To the display 73, the output of the switch 72, which is switched so as to be displayed in the order of the reproduced images, not in the order of decoding, is supplied. As a result, an image is obtained.

As described above, for example, as shown in FIG.
In such a case, when one frame is viewed, there is movement between fields, so such a portion becomes a comb-shaped KS.
According to the present embodiment, such a moving part is coded in the field orthogonal transform mode, so that it can be processed as a field-free blur-free image. High quality video playback. That is, as shown in FIG. 20, for example, at the time of an odd cycle, a moving part is processed in the field motion prediction mode, and a stationary part is processed in the frame orthogonal transform mode. Note that a portion where an image has already been formed in the even cycle is a portion indicated by oblique lines in FIG. In FIG. 21, portions other than the hatched portions, that is, moving portions, are decoded by motion compensation.

In the present embodiment, since only the macroblock processed in the field orthogonal transform mode is decoded in the even cycle, it is necessary to know the macroblock address. There are two methods of knowing the macroblock address. One is the method of transmitting the macroblock address for each of the even-cycle macroblocks described above, and the other is one-field field orthogonal in an odd cycle. This is a method in which information of the conversion mode / frame orthogonal conversion mode is stored, and the address of a macroblock in the field orthogonal conversion mode is converted from a column of each orthogonal conversion mode. The former advantage is that no additional memory is required, and the latter advantage is that transmission information does not increase. Similarly, the quantization width can be realized by transmitting each macroblock without using the method of storing one field in the odd cycle described above.

As described above, according to the decoding apparatus of this embodiment, the frame orthogonal transform mode and the field orthogonal transform mode are switched in macroblock units according to the first and second encoding processing modes. At the same time, switching between the frame motion prediction mode and the field motion prediction mode is performed. In the frame processing, both the odd field and the even field are decoded. In the field processing, only the odd field is decoded, and the quantization width in this cycle is stored. In the next even cycle, the stored information is used to decode the reproduced image by motion-compensating only the macroblock in the field orthogonal transform mode, so that efficient encoded data is transmitted. Can be. That is, a high-quality moving image can be reproduced with a small amount of transmission information.

[0142]

As described above, according to the high-efficiency image signal encoding apparatus of the present invention, for a moving image having a field structure, an image having a small amount of motion, an image having a large amount of motion, and an image in which both are mixed are described. However, field processing or frame processing can be performed efficiently,
Therefore, a high-quality moving image can be reproduced with a small amount of transmission information at the time of decoding by the high-efficiency decoding apparatus of the present invention later.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a schematic configuration of a high-efficiency image signal encoding apparatus according to a first embodiment.

FIG. 2 is a diagram for describing encoding according to the present embodiment.

FIG. 3 is a diagram showing a macro block.

FIG. 4 is a diagram showing a macroblock in a frame orthogonal transform mode.

FIG. 5 is a diagram showing a macroblock in a field orthogonal transform mode.

FIG. 6 is a diagram illustrating motion prediction in the case of a frame orthogonal transformation mode in the first encoding processing mode.

FIG. 7 is a diagram illustrating motion prediction in the case of a field orthogonal transform mode in the first encoding processing mode.

FIG. 8 is a diagram illustrating motion prediction in the case of a second encoding processing mode.

FIG. 9 is a diagram showing a unit block of DCT processing in a frame orthogonal transform mode / field orthogonal transform mode in a specific example format of a digital VTR.

FIG. 10 is a diagram illustrating a state of motion prediction in a first encoding processing mode.

FIG. 11 is a diagram illustrating a state of motion prediction in the present embodiment.

FIG. 12 is a diagram illustrating a state of motion prediction in a second encoding processing mode.

FIG. 13 is a diagram showing a unit block of DCT processing in a frame orthogonal transform mode / field orthogonal transform mode in another specific format of the digital VTR.

FIG. 14 is a diagram showing a set of macro blocks.

FIG. 15 is a diagram for explaining a state of processing in the frame orthogonal transform mode in the example of FIG. 14;

FIG. 16 is a diagram for explaining a state of processing in the field orthogonal transform mode in the example of FIG. 14;

FIG. 17 is a modified example of extension bit addition (pre-prediction).
It is a figure for explaining.

FIG. 18 is a block circuit diagram illustrating a schematic configuration of an encoding device according to a second embodiment.

FIG. 19 is a block diagram illustrating a configuration of a decoder.

FIG. 20 is a diagram showing an image of an odd cycle.

FIG. 21 is a diagram showing an image of an even cycle.

FIG. 22 is a diagram for explaining each prediction image.

FIG. 23 is a diagram showing a data structure.

FIG. 24 is a diagram showing an image of a moving object.

[Explanation of symbols]

10 Frame memory group 12 Difference detector 13 DCT circuit 14 Quantization Unit 15 .... Variable length coding circuit 16 ... Buffer 17 ... Inverse quantizer 18 ... · Inverse DCT circuit 20 ······ Frame memory group with motion compensator 21 ······ Field motion detection circuit 22 ······ Frame motion detection circuit 24 ... Selector 33 ... Frame / field mode determination circuit 34 ... Encoding processing mode determination circuit 35 ... ..Address generators

────────────────────────────────────────────────── ─── Continuation of the front page (56) References JP-A-3-1688 (JP, A) JP-A-3-1399083 (JP, A) JP-A-4-108270 (JP, A) 1990 Image coding Symposium (PCSJ90) p. 175-176 1989 IEICE Autumn National Assembly p. 6-54 1991 IEICE Spring National Convention p. 7-64 (58) Field surveyed (Int. Cl. ⁷ , DB name) H04N 5/91-5/956 H04N ^7/ 24-7/68

Claims (2)

(57) [Claims] 1. An image signal high-efficiency encoding apparatus that encodes a macroblock consisting of a two-dimensional array of a plurality of pixels as a unit, comprising: a motion vector between frames and an absolute value difference of each pixel in the macroblock unit. A motion detecting means comprising a means for detecting a sum and a means for detecting a motion vector between fields consisting of an odd or even number of scans of pixels of the frame divided by the macroblock and an absolute value difference sum of each pixel It is determined whether the frame prediction mode in which motion compensation is performed in units of frames in the macroblock or the field prediction mode in which motion compensation is performed in units of fields in the macroblock is more efficient in performing motion compensation. First mode selecting means for selecting an efficient prediction mode; Either of the frame processing mode in which the orthogonal transformation is performed in units of frames in the macroblock and the field processing mode in which the orthogonal transformation is performed by performing the orthogonal transformation in units of the fields of the macroblock are more efficient in performing the orthogonal transformation. A second mode selection unit for determining whether or not it is good using the information output from the motion detection unit and the first mode selection unit, and selecting an efficient blocking mode; and each macro in one frame. A first encoding processing mode for adaptively switching the blocking into the frame processing mode or the field processing mode for each block, and encoding each macroblock based on each mode; and all macroblocks in one frame. Is performed in the field processing mode, and Only one frame of the odd field in the macroblock is encoded for one frame in the odd cycle of the scanning of the odd field, and then one frame of the even field in the macroblock is encoded for the even cycle of the scanning of the even field in the interlace. A third mode selecting means for determining which of the second encoding processing modes is more efficient in encoding, and selecting an efficient encoding processing mode; and determining whether the odd cycle or the even cycle is selected. Recognize, when the encoding processing mode is the first encoding processing mode, control the frame memory group so as to output a macroblock that has been blocked corresponding to the blocking mode in the odd cycle, When the encoding processing mode is the second encoding processing mode, the odd number Address generation means for controlling a group of frame memories so as to output macroblocks which are divided into blocks corresponding to the field processing mode in the cycle and even number cycles; and motion prediction mode information selected by the first mode selection means. And a motion compensator for receiving the block mode information selected by the second mode selector and executing a motion compensation frame or inter-field prediction in accordance with the mode information. High efficiency coding device. 2. The apparatus receives and decodes reproduced image encoded data and header information including detected motion vector information, motion prediction mode information, blocking mode information, and encoding processing mode information. Inverse variable length encoding means for outputting the detected motion vector information, the motion prediction mode information, the blocking mode information, and the encoding mode information together with the decoded image data; and the image decoded data from the encoding mode information. An address increment value in a frame buffer that accumulates, a head address of each macro block is obtained, an address generating means for giving the head address to the frame buffer, and a relative address of the macro block other than the head address Is added to the frame buffer to access the data, and the detected motion vector is Receiving the frame information, the motion prediction mode information, the blocking mode information, and the encoding processing mode information, performing motion compensation corresponding to the mode information, and sending a motion compensated image signal to the frame buffer. And a motion compensation means configured as described above.