JP4520442B2 - Moving image re-encoding method, moving image re-encoding device, and moving image re-encoding program - Google Patents

Description

  In the present invention, encoded data of the first standard represented by the MPEG2 standard is converted to H.264. The present invention relates to a moving image re-encoding method for converting into encoded data of the H.264 standard or a second standard inheriting the standard with a low amount of computation.

  When re-encoding a stream that has already been encoded according to the MPEG2 standard (ISO / IEC 13818-2) in accordance with the MPEG2 standard, the macroblock type (intra / inter encoding mode, etc.) It is also known that the motion search calculation amount can be reduced if the motion vector is the same value as the input stream (see Non-Patent Document 1).

  FIG. 6 shows an example of a conventional MPEG-2 stream input / MPEG2 stream output moving image re-encoding device. In FIG. 6, the macroblock type / motion vector obtained by the MPEG2 encoding information extraction unit 102 is directly input to the MPEG2 encoding unit 103. Since the stream input to the MPEG2 decoding unit 101 and the stream output from the MPEG2 encoding unit 103 both conform to the MPEG2 standard, all encoding modes and motion vectors selected in the MPEG2 input stream are included. Can also be selected in the MPEG2 output stream. For this reason, the MPEG2 encoding unit 103 simply re-executes the macroblock type (intra / inter encoding mode, etc.) and the motion vector input from the MPEG2 encoded information extraction unit 102 without performing any conversion. It is possible to use.

  On the other hand, the MPEG2 encoding system and the H.264 standard. In the case of H.264 encoding, especially when operating with interlaced video signals as input, the processing unit for selecting a field / frame for each macroblock, the size of a rectangular block for motion compensation, etc. There are the following major differences.

[H. H.264 and MPEG2 (interlaced input / frame coding)]
(1) Unit for performing selection processing for field / frame motion compensation H.264: Macroblock pair MPEG2: Macroblock (2) Field frame DCT processing H.264: The determination of the macroblock pair is inherited and becomes the same. MPEG2: Macroblock (3) Motion compensation unit 264: 16 × 16, 16 × 8, 8 × 16, 8 × 8, 8 × 4, 4 × 8, 4 × 4
MPEG2: 16 × 16
(4) Intra / inter macroblock decision unit H.264: Macroblock MPEG2: Macroblock For this reason, a stream encoded with MPEG2 is converted to H.264. In the case of re-encoding in accordance with the H.264 standard, the decoded image obtained by decoding the MPEG2 stream is designated as H.264. In general, a H.264 encoder (see Non-Patent Document 2) simply performs another encoding.

  FIG. 7 shows MPEG2 decoding of a conventional MPEG2 input stream. The example of the moving image re-encoding apparatus in the case of performing simple re-encoding by a H.264 encoding part is shown. A conventional MPEG2 input stream is converted to H.264. In the moving image re-encoding device 200 for converting to an H.264 output stream, the MPEG2 decoding unit 201 decodes the MPEG2 input stream, The H.264 encoding unit 202 re-encodes the H.264 signal. H.264 output stream is output. In this case, H. The amount of calculation in the H.264 encoding unit 202 is the normal H.264. An amount of calculation equivalent to that of H.264 encoding is required.

  Further, as a technique for converting an image stream by inputting an image stream and re-encoding the moving image, for example, Patent Document 1 (image information conversion apparatus and image information conversion method) described below, Patent Document 2 (Image Information conversion apparatus and method, and recording medium), Patent Document 3 (Image information conversion apparatus and method), Patent Document 4 (Image information conversion apparatus and method), Patent Document 5 (Code amount control apparatus and method, and Image information conversion) Apparatus and method), and Patent Document 6 (image stream conversion apparatus).

However, those described in Patent Documents 1 to 6 are mainly intended to reduce the bit rate between the same types of encoding methods, and are basically the same as those in the prior art shown in FIG. Is.
Nishimura et al., “Development of Real-time MPEG-2 Video Transcoder Software”, Information Processing Research Report, AVM-25-5, 25-30 (1999) H.264 (ITU-T Recommendation H.264 (2003)), “Advanced Video Coding for generic audiovisual services” | ISO / IEC 14496-10: 2003, “Information technology, Coding of audio-visual objects-Part10: Advancedvideo coding ” JP 2000-236547 A JP 2001-128175 A JP 2001-145108 A JP 2001-145113 A JP 2002-209215 A JP 2006-86863 A

  A stream encoded according to the MPEG2 standard is H.264. When re-encoding according to the H.264 standard, If the macroblock type (intra / inter coding mode, etc.), motion vector, etc. at the time of MPEG2 encoding is extracted and the same value as the input MPEG2 stream can be reused, It can be inferred that the calculation amount of H.264 encoding can be reduced. However, especially when interlaced video signals are input, there are significant differences between the two encoding methods, such as the processing unit for selecting a field / frame for each macroblock and the size of a rectangular block for motion compensation. For this reason, the encoding information in the MPEG2 stream is changed to H.264. H.264 cannot simply be reused, and conventionally, a method for reusing MPEG2 encoded information has not been studied.

  For this reason, an MPEG-2 encoded stream is converted to H.264. When converting to an H.264 stream, the decoded image obtained by decoding the MPEG2 input stream is converted to an H.264 stream. The H.264 encoder needs to re-encode in the same manner as a normal input image, and a large amount of calculation is required as compared with the conversion from MPEG2 to MPEG2.

  The present invention encodes encoded data encoded in accordance with a first encoding standard (hereinafter referred to as a first standard) into an encoding in accordance with a second encoding standard (hereinafter referred to as a second standard). The present invention relates to a moving image re-encoding method for converting data into data with a low calculation amount.

  In the following, the MPEG2 standard as the first standard and the H.264 standard as the second standard. The H.264 standard will be described as an example. The present invention can be similarly applied to any coding standard having the same characteristics as H.264. For example, the second standard is H.264. The present invention can be similarly applied even to an encoding standard that inherits or improves the H.264 standard.

  In the first standard represented by at least the MPEG2 standard, the unit for performing the field / frame motion compensation selection process at the time of frame coding of the interlaced video signal is a macro block, and the size of the rectangular block for motion compensation is fixed. H. is an encoding standard. In the second standard represented by the H.264 standard, the unit for performing the field / frame motion compensation selection process at the time of frame coding of the interlaced video signal is a macroblock pair, and the size of the rectangular block for motion compensation is variable. It is a certain coding standard.

The outline of the present invention for solving the above-described problems is as follows.
A bit stream encoded in accordance with the MPEG2 standard is To a stream conforming to the H.264 standard.
・ H. The picture coding structure at the time of H.264 coding is matched with the input MPEG2 stream. Specifically, when the MPEG2 input stream is frame-encoded, H.264 is used. H.264 is also frame-encoded (+ MBAFF encoding mode).
In MBAFF encoding, the mode determined for each macroblock pair is determined from the encoding mode (inter / intra) of two corresponding macroblocks in the MPEG2 input stream.
・ H. The motion vector of the macroblock pair to be re-encoded by H.264 is also determined from the encoding modes of the corresponding two macroblocks in the MPEG2 input stream.

  That is, the present invention performs encoding method conversion processing (transcoding) on the basis of pre-encoding information extracted from the pre-stage “MPEG2 stream” to generate “H.264 stream”. The main feature is that the amount of calculation is reduced by selecting the encoding type and motion vector of the subsequent macroblock pair from the information extracted from the “MPEG2 stream”.

  In the present invention, various types of encoded information obtained at the time of decoding an MPEG2 input stream are designated as H.264. This defines a method for converting to information that can be reused by the H.264 encoding unit.

More specifically, the present invention relates to an MPEG2 bit stream encoded in accordance with the MPEG2 standard (ISO / IEC 13818-2), H.264, and H.264. When converting to a bit stream conforming to the H.264 standard (ITU-T H.264), the MPEG2 decoding unit, the MPEG2 encoded information extraction unit, and the H.264 standard 264 re-encoding unit, the MPEG2 decoding unit decoding the MPEG2 bitstream in accordance with the MPEG2 standard to generate an MPEG2 decoded image, and the MPEG2 encoded information extracting unit extracting the MPEG2 bit A step of extracting MPEG2 encoded information from the encoded bit string described in the stream, and the MPEG2 decoded image and the MPEG2 encoded information are converted into the H.264 format. H.264 re-encoding unit and referring to the MPEG2 encoding information, A re-encoding method in accordance with the H.264 standard, in particular, the H.264 standard, H.264 re-encoding unit converts the MPEG2 decoded image to H.264 format. When re-encoding according to the H.264 standard, if the coding structure of a picture in the MPEG2 coding information is described as a frame structure, the H.264 of the picture is described. The frame picture structure is selected as the picture coding structure at the time of H.264 re-encoding and H.264 re-encoding, and H.264 of the picture. When selecting an MBAFF (macroblock adaptive frame field coding) method for each macroblock pair at the time of H.264 re-encoding, MPEG2 encoding for two corresponding macroblocks at the same position in the picture as the macroblock pair A step of extracting information, and MPEG2 intra-macroblock / inter-macroblock coding selection information, field-frame discrete cosine transform selection information, and inter-macroblock among MPEG2 coding information of the corresponding two macroblocks Based on the field / frame motion compensation selection information written only when encoding is selected, either the frame macroblock pair or the field macroblock pair is selected as the re-encoding method of the macroblock pair. To select And characterized in that it has a and-up.
In particular, according to the present invention, when selecting the MBAFF system for each macroblock pair, none of the extracted MPEG2 encoding information for the corresponding two macroblocks shows field motion compensation, and either one is a frame. If motion compensation is indicated, a frame macroblock pair is selected as the MBAFF system for the macroblock pair, and H.264 is selected. H.264 re-encoding.
In the present invention, the upper macroblock of the MPEG2 coding information for the corresponding two macroblocks indicates intra macroblock coding, and the lower macroblock of the MPEG2 coding information is intermacroblock coding. If the frame motion compensation is indicated, the macroblock pair is set as a frame macroblock pair. H.264 When re-encoding, the step of selecting intra macroblock encoding as the encoding mode of the upper macroblock in the frame macroblock pair, and the encoding mode of the lower macroblock in the frame macroblock pair And the frame motion vector of the lower macroblock in the MPEG2 encoding information for the corresponding two macroblocks is set as the H. of the lower macroblock in the frame macroblock pair. And selecting as H.264 motion vectors.
In the present invention, the lower macroblock of the MPEG2 encoded information for the corresponding two macroblocks indicates intra macroblock encoding, and the upper macroblock of the MPEG2 encoded information indicates inter macroblock encoding and When frame motion compensation is indicated, the macroblock pair is set as a frame macroblock pair. H.264 When re-encoding, a step of selecting intra macroblock coding as a coding mode of a lower macroblock in the frame macroblock pair, and a coding mode of an upper macroblock in the frame macroblock pair Inter macroblock coding is selected, and the frame motion vector of the upper macroblock in the MPEG2 coding information for the corresponding two macroblocks is set as the H.M of the upper macroblock in the frame macroblock pair. And selecting as H.264 motion vectors.

  An MPEG2 bit stream encoded in accordance with the MPEG2 standard (ISO / IEC 13818-2) using the above moving image re-encoding method is converted into an H.264 format. The H.264 standard (ITU-T H.264) is converted into a bit stream that is compliant with the H.264 standard. It is not necessary to perform motion search at the time of H.264 re-encoding, and the amount of calculation can be greatly reduced. In addition, H.264 obtained from the MPEG2 stream by this method. By re-searching a very narrow area in the vicinity of the motion vectors of various horizontal 16 / vertical 16 pixel blocks or the motion vectors of horizontal 16 / vertical 8 pixel blocks used in H.264 re-encoding, the calculation amount is greatly increased. The image quality can be improved.

  As described above, according to the present invention, the macroblock pair type, motion vector, etc., based on the MPEG2 encoded information. By selecting H.264 re-encoding information, The amount of computation required for H.264 re-encoding can be reduced. Further, by re-searching a very narrow area in the vicinity, the image quality can be further improved by slightly increasing the amount of calculation.

  Examples of the present invention will be described below. FIG. 1 shows a configuration example of a moving image re-encoding device according to the present invention. The moving image re-encoding device 10 includes an MPEG2 decoding unit 11, an MPEG2 encoded information extraction unit 12, H.264, and the like. The H.264 re-encoding unit 13 is provided. H. H.264 re-encoding unit 13 includes MPEG2 encoded information interpretation unit 14 and H.264 re-encoding unit 13. An H.264 encoding unit 15 is provided.

An MPEG2 input stream encoded in accordance with the MPEG2 standard is input to the MPEG2 decoding unit 11 and decoded in accordance with the MPEG2 standard, and an MPEG2 decoded image is output from the MPEG2 decoding unit 11. This MPEG2 input stream is also input to the MPEG2 encoded information extracting unit 12, and the MPEG2 encoded information is extracted by the MPEG2 encoded information extracting unit 12. Specifically, the following information is extracted as the MPEG2 encoded information. Hereinafter, the macroblock may be abbreviated as MB.
(1) Picture coding structure (2) Picture type of each picture (3) Macroblock type of each macroblock (intra MB / inter MB selection information)
(4) DCT type of each macroblock (field DCT / frame DCT selection information)
(5) MC type of each macroblock (field / frame motion compensation selection information)
(6) Motion vector of each macroblock The extracted MPEG2 encoded information is input to the MPEG2 encoded information interpretation unit 14. The MPEG2 encoded information interpretation unit 14 uses the four types of encoded information shown in the MPEG2 encoded information to generate H.264 encoding information. The following four H.264 codes used in H.264 encoding: H.264 re-encoding information is calculated.
(1) Picture type of each picture (2) Field / frame pair type of macroblock pair (3) Predicted block size of each macroblock within macroblock pair (4) Prediction within each macroblock within macroblock pair Block Motion Vector FIG. 2 shows H.264 in the MPEG2 encoded information interpretation unit 14. The flow of the calculation method of H.264 re-encoding information is shown. The frame structure determination unit 141 determines whether the picture coding structure in the MPEG2 coding information input to the MPEG2 coding information interpretation unit 14 has a frame structure. If the flag indicating the frame structure is not shown in the MPEG2 input stream and the frame structure determination unit 141 determines that the frame structure is not the frame structure, the MPEG2 / H. The operation of the H.264 encoded information conversion unit 142 is not defined in the present invention. This is because the present invention is applied only when the input MPEG2 stream has a frame structure. In the following description, only the case where the MPEG2 stream has a frame structure will be described. The picture type in the MPEG2 encoded information input to the MPEG2 encoded information interpretation unit 14 is the same as the H.264 standard. It is output as a picture type of H.264 re-encoding information.

  The macroblock type, DCT type, MC type, and motion vector in the MPEG2 encoded information input to the MPEG2 encoded information interpretation unit 14 are in accordance with the processing flow shown in FIG. H.264 re-encoded information.

  In addition, FIG. 4 shows an arrangement in which each flow is classified into b-1 to b-6 patterns. Two macroblocks indicated by squares in FIG. 4 indicate MPEG2 encoding information for two corresponding macroblocks. For each pattern (b-1 to b-6) in FIG. 4, the type (field or frame) of the selected macroblock pair and the motion vector in the case of an inter macroblock will be described in detail below. The item [Axx] in the following description is H.264. In the H.264 re-encoding, a macroblock pair, [Bxxx] or [Bxx] is H.264. 2 is a method for extracting a motion vector of each prediction block in H.264 re-encoding.

[B-1 pattern]
[A11] Top field MB of field macroblock pair.
[B111] The top field vector (or frame vector if not present) of the upper MB of the corresponding two macroblocks in the MPEG2 encoded information H.264 of the upper horizontal 16-pixel vertical 8-pixel prediction block of the top field macroblock pair in the H.264 field macroblock pair. Select as H.264 motion vector. [B112] The top field vector (or frame vector if not) of the lower MB of the corresponding two macroblocks in the MPEG2 encoded information is set to H.264. H.264 of the lower horizontal 16 pixel vertical 8 pixel prediction block in the top field macroblock pair in the H.264 field macroblock pair. Select as H.264 motion vector.
[A12] The bottom field MB of the field macroblock pair.
[B121] The bottom field vector (or frame vector if not present) of the upper MB of the corresponding two macroblocks in the MPEG2 encoded information H.264 of the upper horizontal 16-pixel vertical 8-pixel prediction block of the bottom field macroblock pair in the H.264 field macroblock pair. Select as H.264 motion vector. [B122] The bottom field vector (or frame vector if not present) of the lower MB of the corresponding two macroblocks in the MPEG2 encoded information H.264 of the lower horizontal 16 pixel vertical 8 pixel prediction block in the bottom field macroblock pair in the H.264 field macroblock pair. Select as H.264 motion vector.

[B-2 pattern]
[A21] Top field MB of field macroblock pair.
[B21] Among the two corresponding macroblocks in the MPEG2 encoded information, the top field vector of the field-predicted MB is defined as H.264. H.264 of the top field macroblock pair (16 × 16) in the H.264 field macroblock pair. Select as H.264 motion vector.
[A22] Bottom field MB of field macroblock pair.
[B22] Of the two corresponding macroblocks in the MPEG2 encoded information, the bottom field vector of the field-predicted MB is defined as H.264. H.264 of bottom field macroblock pairs (16 × 16) in H.264 field macroblock pairs. Select as H.264 motion vector.

[Pattern b-3]
[A31] Upper MB of frame macroblock pair.
[B31] Of the two corresponding macroblocks in the MPEG2 encoded information, the frame motion vector of the upper macroblock is set to H.264. H.264 of the upper macroblock in the H.264 frame macroblock pair. Select as H.264 motion vector.
[A32] Lower MB of frame macroblock pair.
[B32] Of the two corresponding macroblocks in the MPEG2 encoded information, the frame motion vector of the lower macroblock is set to H.264. H.264 of the lower macroblock in the H.264 frame macroblock pair. Select as H.264 motion vector.

[Pattern b-4]
[A41] Upper MB of frame macroblock pair.
[B41] Of the two corresponding macroblocks in the MPEG2 encoded information, if the upper macroblock is frame motion compensation, the H.264 also selects inter MB, and uses the motion vector for frame motion compensation as H.264. Select as H.264 motion vector. In the case of intra MB instead of frame motion compensation, H.264 is used. H.264 also selects intra MB.
[A42] Lower MB of frame macroblock pair.
[B42] Of the two corresponding macroblocks in the MPEG2 encoded information, if the lower macroblock is frame motion compensation, the H.264 also selects inter MB, and uses the motion vector for frame motion compensation as H.264. Select as H.264 motion vector. In the case of intra MB instead of frame motion compensation, H.264 is used. H.264 also selects intra MB.

  These extraction processes will be described according to the macroblock pair type / macroblock type / predicted block type extraction flow shown in FIG.

  First, in the case of MPEG2 encoding information for two corresponding macroblocks, at least one of them indicates field motion compensation and both are selected for inter MB encoding (summarizing this, b-1 in FIG. The field macroblock pair is selected as the corresponding macroblock pair, and the two macroblocks in the macroblock pair are all selected as inter MB coding, and the horizontal 16 pixels and the vertical 8 pixels. A prediction block type is selected. For the four horizontal 16-pixel vertical 8-pixel prediction blocks in the macroblock pair, based on the MPEG2 encoding information for the two macroblocks, the motion vector is calculated according to the extraction method of [b-1 pattern] described above. select.

  Similarly, in the MPEG2 encoding information for two corresponding macroblocks, at least one indicates field motion compensation and one of them is an intra MB (in summary, the combination of b-2 in FIG. The field macroblock pair is selected as the corresponding macroblock pair, the inter-MB coding is selected for both of the two macroblocks in the macroblock pair, and a prediction block of 16 horizontal pixels and 16 vertical pixels is selected. A type is selected. For two macroblocks in a macroblock pair, a motion vector is selected according to the extraction method of [b-2 pattern] described above based on the MPEG2 encoding information for the two macroblocks.

  Similarly, neither field motion compensation is shown, and both are inter MB and frame motion compensation (summarizing this will represent the combination of b-3 in FIG. 4), the corresponding macroblock pair A frame macroblock pair is selected, inter-MB coding is selected for both of the two macroblocks in the macroblock pair, and a prediction block type of 16 horizontal pixels and 16 vertical pixels is selected. For two macroblocks in a macroblock pair, a motion vector is selected according to the extraction method of [b-3 pattern] described above based on the MPEG2 encoding information for the two macroblocks.

  Similarly, none of them shows field motion compensation, and at least one is inter MB and frame motion compensation, and the other is an intra macroblock (summarizing this, the combination of b-4 in FIG. The corresponding macroblock pair is selected as a frame macroblock pair, two macroblocks in the macroblock pair are selected as inter MB coding, and as the other is intra MB coding, A prediction block type of 16 pixels vertical 16 pixels is selected. For two macroblocks in a macroblock pair, a motion vector is selected or intra MB according to the extraction method of [b-4 pattern] described above based on the MPEG2 encoding information for the two macroblocks. Select.

  Similarly, in the case where both are intra MBs and one of them is a field discrete cosine transform (DCT) (summarizing this, it represents the combination of b-5 in FIG. 4), the corresponding macroblock pair is a field. A macroblock pair is selected.

  Similarly, in the case of both intra MBs and frame discrete cosine transform (DCT) (summarizing this, the combination of b-6 in FIG. 4 is represented), the corresponding macroblock pair is the frame macroblock pair. Is selected.

  MPEG2 decoded image and H.264 H.264 re-encoding information is H.264. It is input to the H.264 encoding unit 15. H. The H.264 encoding unit 15 uses the MPEG2 decoded image as an input video signal to be encoded. For various types of encoded information shown in the H.264 re-encoded information, see H.264. Without recalculation in the H.264 encoding unit 15. H.264 re-encoded information is used as it is, and the H.264 re-encoded information is used as it is. It is possible to reduce the amount of computation required for H.264 encoding and at the same time suppress the deterioration in image quality.

  From the encoded information obtained from the MPEG2 input stream by the above moving image re-encoding method, It is possible to calculate the field / frame macroblock pair type at the time of H.264 re-encoding, the macroblock type / predicted block size / motion vector of the predicted block of each macroblock in the macroblock pair without re-searching. It becomes.

  Also, the moving image re-encoding method described below searches a wide search range as in the case of simple re-encoding by the conventional method by re-searching the pixels near the motion vector of the calculated prediction block. Without this, the image quality is improved with an additional amount of computation only for the neighborhood search.

  FIG. 5 shows a configuration example of another moving image re-encoding device according to the present invention. MPEG2 decoding unit 21, MPEG2 encoded information extraction unit 22, MPEG2 encoded information interpretation unit 24, H.264 in the moving image re-encoding device 20 of FIG. The H.264 encoding unit 25 has the same function as the block of the same name in the moving image re-encoding device 10 shown in FIG. In the moving image re-encoding device 20 of FIG. The H.264 re-encoding unit 23 is different from the moving image re-encoding device 10 shown in FIG.

  An MPEG2 input stream encoded according to the MPEG2 standard is input to the MPEG2 decoding unit 21 and decoded according to the MPEG2 standard, and an MPEG2 decoded image is output from the MPEG2 decoding unit 21. This MPEG2 input stream is also input to the MPEG2 encoded information extraction unit 22, and MPEG2 encoded information is extracted and input to the MPEG2 encoded information interpretation unit 24. These processes are the same as the processes of the moving picture re-encoding device 10 of FIG. Further, the MPEG2 encoded information interpretation unit 24 performs various H.264 encoding based on the MPEG2 encoded information. The method for obtaining the H.264 re-encoding information is also equivalent to the example described above.

  In this embodiment, the motion vector of each prediction block obtained by the MPEG2 encoded information interpretation unit 24 is input to the neighborhood search unit 26. The neighborhood search unit 26 uses the MPEG2 decoded image input from the MPEG2 decoding unit 21 as a reference image, and re-searches the number of neighborhood pixels of the image position in the reference image indicated by the motion vector of each prediction block. The neighborhood pixel range to be searched again is H.264. It may be determined according to the calculation performance of the H.264 re-encoding unit 23, that is, the calculation amount per unit time that can be executed. In general, it may be about 1 to 4 pixels with integer pixel accuracy. A re-search with half-pixel accuracy and quarter-pixel accuracy peculiar to H.264 encoding may be performed.

  The motion vector after the neighborhood re-search obtained by the neighborhood search unit 26 is H.264. H.264 encoding unit 25, the motion vector after the neighborhood re-search is obtained as H.264. H.264 as a motion vector in H.264 re-encoding. H.264 re-encoding is performed. The MPEG2 encoded information interpretation unit 24 sends the H.264 information. H.264 encoding unit 25 outputs H.264. For H.264 re-encoding information, the motion vector for each prediction block is not used. Other information included in the H.264 re-encoding information, specifically, the picture type of each picture, the field / frame pair type information of the macroblock pair, and the predicted block size of each macroblock in the macroblock pair are described in H.264. H.264 encoding unit 25 does not recalculate and uses it for re-encoding as it is. It is possible to reduce the amount of computation required for H.264 encoding and at the same time suppress the deterioration in image quality.

  The above moving image re-encoding process can be realized not only by hardware and firmware but also by a computer and a software program, and the program can be recorded on a computer-readable recording medium and provided. It can also be provided through a network.

It is a figure which shows the structural example of the moving image re-encoding apparatus which concerns on this invention. It is the schematic of an MPEG2 encoding information interpretation part. It is MB pair type, macroblock type, and prediction block type extraction flowchart. It is explanatory drawing of MB pair type, macroblock type, and prediction block type extraction. It is a figure which shows the structural example of the other moving image re-encoding apparatus which concerns on this invention. It is a figure which shows the example of the moving image re-encoding apparatus which performs the conventional MPEG2 stream input and MPEG2 stream output. Conventional MPEG2 stream input. 1 is a diagram illustrating an example of a moving image re-encoding device that performs H.264 stream output.

Explanation of symbols

10, 20 Moving picture re-encoding device 11, 21 MPEG decoding unit 12, 22 MPEG2 encoded information extraction unit 13, 23 H.264 re-encoding unit 14, 24 MPEG2 encoded information interpretation unit 15, 25 263 encoding unit 26 neighborhood searching unit

Claims (3)

Conforms to the first standard, which is a coding block in which the unit for performing field / frame motion compensation selection processing is a macroblock and the size of a rectangular block for motion compensation is fixed when interlaced video signals are encoded. A first standard decoding unit that decodes the encoded input bitstream to generate a first standard decoded image;
A first standard encoded information extracting unit that extracts first standard encoded information from an encoded bit string described in the input bit stream;
The first standard decoded image and the first standard encoded information are input, and the first standard decoded image is subjected to field encoding when interlaced video signals are frame-coded while referring to the first standard encoded information. The second standard re-encoding re-encoding in accordance with the second standard which is a coding standard in which the unit for performing the frame motion compensation selection process is a macro block pair and the size of the rectangular block for motion compensation is variable And
A moving image re-encoding method for converting an input bit stream of a first standard into a bit stream compliant with a second standard,
When the second standard re-encoding unit re-encodes the first standard decoded image according to the second standard, a coding structure of a picture in the first standard encoded information is a frame structure. If so, the step of selecting a frame picture structure as a picture coding structure at the time of the second standard re-encoding of the picture and performing re-encoding according to the second standard;
When selecting a macroblock adaptive frame field coding (hereinafter referred to as MBAFF) method for each macroblock pair during the second standard re-encoding of the picture,
Extracting first standard coding information for two corresponding macroblocks at the same position in the picture as the macroblock pair;
Among the first standard coding information for the corresponding two macroblocks, selection information for first standard intra macroblock / intermacroblock coding, selection information for field / frame discrete cosine transform, and intermacroblock coding A step of selecting either a frame macroblock pair or a field macroblock pair as a re-encoding method of the macroblock pair based on selection information of field / frame motion compensation written only when it is selected It has a door,
In selecting the MBAFF method for each macroblock pair,
In the extracted first standard encoded information for two corresponding macroblocks, if neither indicates field motion compensation and one indicates frame motion compensation, the MBAFF of the macroblock pair Select a frame macroblock pair as the method and perform re-encoding in accordance with the second standard.
In addition, one of the upper and lower macroblocks of the first standard encoded information for the corresponding two macroblocks indicates intra macroblock encoding, and the upper or lower side of the first standard encoded information. If the other macroblock of is indicating inter-macroblock coding and frame motion compensation,
In re-encoding the macroblock pair as a frame macroblock pair according to the second standard,
Selecting intra macroblock encoding as the encoding mode of the one-side macroblock in the frame macroblock pair;
Inter macroblock coding is selected as the coding mode of the other macroblock in the frame macroblock pair, and the other side of the first standard coding information for the corresponding two macroblocks is selected. Selecting a frame motion vector of the macroblock as a second standard motion vector of the other macroblock in the frame macroblock pair . Conforms to the first standard, which is a coding block in which the unit for performing field / frame motion compensation selection processing is a macroblock and the size of a rectangular block for motion compensation is fixed when interlaced video signals are encoded. A first standard decoding unit that decodes the encoded input bitstream to generate a first standard decoded image;
A first standard encoded information extracting unit that extracts first standard encoded information from an encoded bit string described in the input bit stream;
The first standard decoded image and the first standard encoded information are input, and the first standard decoded image is subjected to field encoding when interlaced video signals are frame-coded while referring to the first standard encoded information. The second standard re-encoding re-encoding in accordance with the second standard which is a coding standard in which the unit for performing the frame motion compensation selection process is a macro block pair and the size of the rectangular block for motion compensation is variable And
A moving image re-encoding device for converting an input bit stream of a first standard into a bit stream compliant with a second standard,
The second standard re-encoding unit includes:
When re-encoding the first standard decoded image in accordance with the second standard, if the coding structure of a picture in the first standard coding information is described as a frame structure, the picture Means for selecting a frame picture structure as a picture coding structure at the time of re-encoding of the second standard and performing re-encoding according to the second standard;
In selecting the MBAFF method for each macroblock pair when the picture is second-standard re-encoded,
Means for extracting first standard coding information for two corresponding macroblocks at the same position in the picture as the macroblock pair;
Among the first standard coding information for the corresponding two macroblocks, selection information for first standard intra macroblock / intermacroblock coding, selection information for field / frame discrete cosine transform, and intermacroblock coding Means for selecting either a frame macroblock pair or a field macroblock pair as a re-encoding method of the macroblock pair based on field / frame motion compensation selection information written only when it is selected provided with a door,
The second standard re-encoding unit includes:
In selecting the MBAFF method for each macroblock pair,
In the extracted first standard encoded information for two corresponding macroblocks, if neither indicates field motion compensation and one indicates frame motion compensation, the MBAFF of the macroblock pair Select a frame macroblock pair as the method and perform re-encoding in accordance with the second standard.
The second standard re-encoding unit
Of the first standard coding information for the corresponding two macroblocks, one of the upper and lower macroblocks indicates intra macroblock coding, and the other of the first standard coding information is the upper or lower side. If a macroblock of the above indicates inter-macroblock coding and frame motion compensation,
In re-encoding the macroblock pair as a frame macroblock pair according to the second standard,
Means for selecting intra macroblock coding as the coding mode of the macroblock on the one side in the frame macroblock pair;
Inter macroblock coding is selected as the coding mode of the other macroblock in the frame macroblock pair, and the other side of the first standard coding information for the corresponding two macroblocks is selected. And a means for selecting a frame motion vector of the macro block as a second standard motion vector of the macro block on the other side in the frame macro block pair . A moving image re-encoding program for causing a computer to execute the moving image re-encoding method according to claim 1 .

Images