FR2910216A1 - Macroblock adaptative field/frame coding mode selecting method for coding image part of video sequence, involves respectively selecting frame and image coding modes when counter value is higher and lower than preset threshold - Google Patents

Abstract

The method involves determining absolute departure value between luminous data representing a considered pixel and luminous data representing a pixel group (21) comprising the considered pixel and two pixels. A value of a counter is increased when the absolute value is higher than a preset threshold (SA), and the pixel considering, departure value determining and counter value increasing processes are repeated for each portion of the image. The frame and image coding modes are respectively selected when the counter value is higher than preset threshold (SD) and lower than the threshold (SD). Independent claims are also included for the following: (1) a recording medium comprising computer program comprises a set of instructions for executing a steps of the coding mode selecting method (2) a device for selecting an frame/image coding mode of a part of an image of video sequence (3) a method for coding image according to the macroblock adaptative field/frame (4) a method for encoding MPEG2 compression format to a MPEG4 compression format.

Description

The invention relates to the field of video compression and in particular

  MPEG4 / AVC video compression in an adaptive macro-block image / frame mode (generally referred to as the MBAFF: Adaptive Macroblock Field / Frame mode).

  In particular, the present invention relates to a method of the method of selecting a frame or frame coding mode for encoding at least part of an image of a video sequence, said image portion comprising pixels divided into n rows and m columns. In video compression, images and image sequences are analyzed to define encoding parameters representative of the image data. These coding parameters must be carefully chosen to minimize the cost of coding. When an encoded video uses MPEG4 / AVC encoding mode MBAFF, it is necessary to indicate to the MPEG4 encoder which encoding mode to use from the picture mode or the frame mode to encode the macroblock pairs. In raster mode, the pair of macroblocks consists of two macroblocks and for each of them, the lines are formed from the lines of the same frame. In image mode, macroblocks retain the notion of interlaced image.

  The invention is in an MPEG4 / AVC encoding context using a first encoding in the MPEG2 (transcoding) format. In this case, it is possible to recover some parameters calculated for coding according to the MPEG2 format by decoding the coded video in MPEG2 to encode this same video in an MPEG4 / AVC format. These coding parameters include in particular the motion vectors of the macroblocks of the image and the inter / intra coding choices. The invention also extends to pure MPEG4 / AVC video coding since it describes the choice of use of the MBAFF mode specific to MPEG4 / AVC.

  According to the MPEG2 format, the images are coded by macroblock without identifying an MBAFF encoding mode. The decoding of a coded video 2910216 2 according to the MPEG2 format thus gives no information on the coding mode to be used for each pair of macroblocks of the picture since, unlike the MPEG4 format coding in the MBAFF mode, the coding MPEG2 format does not process image data by pair of macroblocks. To solve this problem, it is possible to arbitrarily choose one of the two image or frame modes and to apply it to all pairs of macroblocks of the image. However, this arbitrary choice is not optimal. Indeed, when the image mode is chosen for a sequence of images having many movements, the quality of the decoded images is unsatisfactory because of the high frequencies generated by the interlaced on the contours. When, on the contrary, the frame mode is chosen for homogeneous image sequences, the coding cost is not optimized.

In particular, in the case of transcoding from an MPEG2 format to an MPEG4 / AVC format, the coding cost is penalized when the frame mode is chosen for P and B type images with MPEG2 frame predictions since each Macroblock is divided into two sub-partitions of 16 rows by 8 columns with motion vectors from each sub-partition. Accordingly, in this case, the choice of the frame mode increases the number of motion vectors to be encoded by two. It is an object of the invention to provide a method for selecting the best mode from the image mode or the frame mode for each pair of macroblocks of an image in order to minimize coding cost while ensuring sufficient quality of the images. images, when encoding in MBAFF mode of MPEG4 format. To this end, the subject of the invention is a selection method of the aforementioned type, characterized in that it comprises the following steps: a) defining a portion of the or each image portion; B) considering a pixel belonging to said image portion; C) determining the absolute value of the difference between a light data representative of the pixel in question and a light data representative of a group of pixels, said group of pixels comprising the pixel in question and at least two pixels belonging to different lines said pixel 5 considered; d) incrementing by one step a value of a counter when said absolute value is greater than a first predefined threshold; e) repeating steps b) to d) for each pixel of the image portion; and f) selecting the frame encoding mode for the or each image portion when the counter value is greater than a second predefined threshold and the image encoding mode when the counter value is less than the second predefined threshold. The encoding mode selected by the selection method according to the invention is intended to be used for coding the macro-block adaptive mode (MBAFF) image portion of the MPEG4 compression format. According to particular embodiments, the selection method comprises one or more of the following characteristics: said luminous datum comprises a luminance datum; The step of determining the absolute value comprises a step of searching for the median value of the light data of the group of pixels; the image portion comprises all the pixels of the image part except the pixels of the first and the last lines, I being a natural number; when the pixel in question belongs to the kth line, the group of pixels comprises the pixel considered and the pixels between the k-th line and the k + lth line; - 1 is equal to 1; The image portion is a pair of macroblocks of said image, said pair of macroblocks comprising pixels distributed along 32 lines and 16 columns; it comprises, before step a) of definition of an image portion, the following steps: - to receive an image; - consider part of the image received; and it comprises, after the selection step g), the following steps: reset the value of the counter and consider another part 10 of said image; and - repeat steps a) to g) for all parts of the image. - the image is an interlaced image. The invention also relates to a method for encoding at least one macro-block adaptive image / frame image of the MPEG4 compression format, the method comprising the steps of the method of selecting a macro-block mode. encoding according to any one of the preceding features, and a step of coding each part of the image according to the selected coding mode. The invention also relates to a recording medium 20 comprising a computer program comprising instructions for performing the steps of the selection method, when the program is executed on a computer. The subject of the invention is also a device for selecting a frame or image encoding mode of at least a part of an image of a video sequence, said part of images comprising pixels distributed along n lines and m columns, the selected coding mode being intended to be used for the coding of the image part according to the macroblock adaptive mode of the MPEG4 compression format, characterized in that it comprises: means for determining the value absolute of the difference between a luminous datum of the pixel in question and a luminous datum representative of a group of pixels, said group of pixels comprising the pixel considered and at least two pixels belonging to different lines of said pixel in question; a counter whose value is adapted to be incremented by one step when the absolute value calculated by the determination means is greater than a first predefined threshold; and selection means adapted to select the frame coding mode for the or each image part when the counter value is greater than a second predefined threshold, and the image coding mode when the counter value is lower than the second one; predefined threshold.

The invention further relates to a method of transcoding the MPEG2 compression format to the MPEG4 compression format. The selection of a frame or frame coding mode for coding at least part of an image of a macroblock adaptive mode (MBAFF) video sequence of the MPEG4 compression format is carried out according to the method of selection mentioned above. The invention will be better understood on reading the description which follows, given solely by way of example and with reference to the drawings, in which: FIG. 1 is a block diagram showing a MPEG2 transcoding system; MPEG4 / AVC incorporating a selection device according to the invention; Figure 2 is a block diagram showing an image to be encoded; and Fig. 3 is a flow chart showing steps of the selection process according to the invention. An MPEG2UMPEG4 / AVC transcoding system is illustrated schematically in FIG. 1. This transcoding system comprises an encoder 1 and a decoder 2 connected to the encoder 1. The encoder 1 is capable of receiving images 3, 4 interleaved with a sequence video 5, to code these images by macroblock according to the MPEG2 format and to transmit them to the decoder 2.

The decoder 2 is able to decode the images 3, 4 and to extract coding parameters specific to each image 3, 4 such as for example the motion vectors and the Infra-Inter coding choices of these images. The transcoding system further comprises a transcoder 6 connected to the decoder 2, an encoder 7 connected to the transcoder 6 and a coding mode selection device 8 according to the invention, connected to the encoder 7. The transcoder 6 is adapted to receive the coding parameters coded according to the MPEG2 format of the decoder 2, converting them to the MPEG4 format and finally transmitting them to the encoder 7. The encoder 7 comprises a processing unit 7A. When the MBAFF mode MPEG4 / AVC format is used, the processing unit 7A is able to receive the images 3, 4 of the video 5 and to arrange the pairs 11, 12 of macroblocks 10 images in a frame mode or image from the encoding mode choices generated by the selection device 8. The encoder 7 is adapted to encode pairs of macroblocks from the processing unit 7A according to the MPEG4 / AVC format MBAFF mode. from the MPEG4 coding parameters received from the transcoder 6. A macroblock 10 comprises pixels distributed along 16 lines and 16 columns. A pair 11 or 12 of macroblocks comprises two adjacent macroblocks 10 arranged vertically one over the other in the direction of the columns. Thus, each pair 11, 12 of macroblocks has 32 rows and 16 columns. As can be seen in FIG. 2, when a pair of macroblocks is arranged according to the image mode, this pair successively comprises a line 13 belonging to the current frame and a line 14 belonging to the next frame. Thus, in this pair 11 of macroblocks, two successive image frames 3, 4 belonging to the same image are interleaved. When a pair of macroblocks is arranged according to the frame mode, all the lines of a macroblock belong to the top or top frame of a current image and all the lines 16 of the other macro. - 2910216 7 block belong to the lower frame or bottom frame of the same picture. The selection device 8 according to the invention is adapted to receive the images 3, 4 of the video 5 before encoding, to analyze them and to provide a choice of coding mode mode among the frame or picture mode. This choice of coding mode is determined for each pair 11, 12 of macro-blocks of the images 3,4. The selection device 8 comprises luminance determination means 17 and calculation means 18 connected to the determination means 17. The determination means 17 comprise, in particular, pixel selection means 19A inside a macro-block and a median filter 19B. The selection means 19A are able to consider in turn a pixel and a group of pixels belonging to a portion 20 of a macroblock 11 and to determine a luminous datum such as the luminance of these pixels. The median filter 19B is adapted to be applied to the group of pixels 21 to determine the median luminance of this pixel group 21 as explained later in the description.

The calculation means 18 are able to calculate the absolute value of the difference between the luminance of the pixel in question and the median luminance of the pixel group 21. This difference represents the difference between the luminance of the pixel considered and the median luminance of the pixel. group of pixels. The selection device 8 further comprises first comparison means 22 connected to the calculation means 18 and a first memory 23 connected to the first comparison means 22. The first comparison means 22 are adapted to compare the absolute value calculated by the calculation means 18 at a first predefined threshold SA recorded in the first memory 23 and generating a positive signal when the absolute value is greater than the first threshold SA.

The selection device 8 further comprises a counter 24 connected to the output of the first comparison means 22, a second memory 25 and second comparison means 26 connected at input to the counter 24 and to the second memory 25, and output to encoder 7.

The counter 24 is able to receive a positive signal from the output of the first comparison means 22 and to increment by one step when receiving such a signal. The counter 24 is also connected to the determination means 17 and is adapted to receive therefrom a warning signal when the last pixel of the portion 20 has been considered. The counter 24 is able to transmit to the second comparison means 26 the incremented value on receipt of this warning signal. The second comparison means 26 are able to compare the value received from the counter 24 with a second predefined threshold SD recorded in the second memory 25 and to select a frame coding mode for the pair of macroblocks considered 11 when the value of the counter 24 is greater than the second predefined threshold SD and the image coding mode when the value of the counter 24 is below this threshold. The method for selecting a frame or frame coding mode according to the invention is illustrated in FIG.

The method according to the invention comprises an initial step 40 during which the portion 20 of the pair 11 of macroblocks of the image 3 is considered. This portion 20 comprises all the pixels of the pair 11 of macroblocks with the exception of the pixels of the first and the last lines.

During a step 42, the pixel x (i, j) of the second line and of the first column is defined as the current pixel, by the determination means 17. During a step 44, the two pixels x (i, j-1), x (i, j + 1) contiguous to the current pixel x (i, j), belonging to the same column as the current pixel x (i, j) and to different lines are defined as being the current pixel group 29. Thus, the pixel group 21 comprises three contiguous pixels aligned vertically. Then, in step 46, the median filter 19B is applied to the pixel group 21 to determine the median of the luminance values of the pixels of the current pixel group. The median filter 19B ranks the luminance values in ascending order and then delivers the value in the middle of the series. For example, when the luminance values of the pixel group 20 are equal to x (i, j-1) = 114; x (i, j) = 110; x (i, j + 1) = 120, the value delivered by the median filter is equal to 114. The determination means 17 then transmit the luminance value of the current pixel and the median luminance value of the pixel group 20 to the pixels. computing means 18. In step 48, the absolute value of the difference between, on the one hand, the luminance value of the current pixel x (i, j) and, on the other hand, the median luminance value of the group Current pixels are calculated by the calculation means 18. In step 50, the absolute value calculated in step 48 is compared with the first predefined threshold SA by the first comparison means 22. The first threshold SA is for example equal to 10. When this value is lower than the first threshold SA, the process goes to step 52 during which the determination means 17 check whether all the pixels of the portion 20 have been considered. When this is not the case, the method proceeds to step 42, during which the pixel of the next line x (i + 1, j) or the pixel of the next column x (i, j + 1) is considered. When the absolute value calculated in step 48 is greater than the first threshold SA, the counter 24 is incremented by one step in a step 54 and the process proceeds to the verification step 52.

Steps 40 to 50 are then repeated until all the pixels of the portion 20 of the macroblock 11 have been considered.

If at step 52 all the pixels of the portion 20 have been considered as current pixels, the detecting means 17 transmits a warning signal to the counter 24 and the value of the counter 24 is compared to the second threshold SD at during a step 56.

When the value of the counter 24 is greater than the second threshold SD, the second comparing means 26 outputs a signal representative of the frame coding mode in step 58. When the value of the counter 24 is less than the second threshold SD, the second comparison means 26 transmit a signal representative of the image coding mode in step 60.

The second threshold SD is for example equal to 48. During a step 62, the counter 24 is reset to 0 and the selection process returns to step 40 during which the next macroblock 12 of image 3 is considered. Since the images 3, 4 of the video 5 are interlaced images, the difference between the median of the pixel group 21 and the pixel in question is small in a homogeneous area from one frame to another. In this case, the frame mode is preferred. On the other hand, when the difference between the median of the group 21 and the pixel in question is large, the interlaced images 2,4 comprise motion and the image mode is preferred.

As a variant, the selection method according to the invention is used in a context of direct MPEG4 coding, that is to say without using the parameters obtained by MPEG2 coding. In a variant, the group of pixels comprises a number of pixels greater than 3.

As a variant, this group of pixels is arranged diagonally. Alternatively, the selection method according to the invention is used for non-interlaced images that is to say progressive. As a variant, the light data is or comprises the chrominance of the pixels of the macroblock.

As a variant, the median filter 19B is a mean filter making it possible to calculate the average value between the luminance values of the group of pixels. Advantageously, the selection method according to the invention makes it possible to obtain a flow rate 10% lower compared to a frame MBAFF mode only for an identical quality. In addition, this selection method allows a gain of 0.85 decibels in quality and 2.3% in flow compared to image mode MBAFF only for a transcoding of an intra image.

The invention also relates to a method for encoding at least one image according to the adaptive frame / frame mode of the macro-block of the MPEG4 compression format. This method comprises the steps of the method of selecting an encoding mode described above, a step of arranging the macroblocks according to the image or frame mode from the modes selected by this method and a coding step of each macro-block. According to one variant, the coding method comprises the following steps: coding of interlaced images 3, 4 according to the MPEG2 compression format; Decoding the coded images 3, 4 to obtain the coding parameters in MPEG2 format relating to the movement fields and to the Inter-Infra choices; - transcoding coding parameters in MPEG4 format; selecting an image or frame coding mode for each macroblock pair of images, according to the selection method mentioned above; arrangement of the macro-blocks of the irnages according to the selected coding modes; and coding in the MPEG4 format of the macroblocks arranged from the coding parameters in the MPEG4 format.

The invention also relates to a recording medium comprising a computer program comprising instructions for executing the steps of the selection method described above when the program is executed on a computer.

When transcoding MPEG2 / MPEG4 and in a particular implementation, a mode selection algorithm determines the encoding mode for an image or for a row of macroblocks of the image (slice). When the MBAFF encoding mode is chosen, for example for an image, the method of selecting the "frame" or "image" encoding mode, as previously described, is applied for all macroblock pairs of the image. The information from the MPEG2 data stream and allocated to each of the macroblocks of the image relating to the pair of macroblocks, for example the motion vectors, are then adapted to the coding mode "frame" or "image" selected for this pair of 15 macroblocks. The invention described relates to the MBAFF mode of the MPEG4 compression format. It is obvious that it can be applied to any type of format or standard exploiting an MBAFF mode similar to that of the MPEG4 format. 20

Claims (13)

  A method of selecting a frame or frame coding mode for encoding at least a portion (11, 12) of an image of a video sequence, said image portion (11, 12) including pixels divided into n rows and m columns, characterized in that it comprises the following steps: - a) defining (40) a portion (20) of the or each image portion (11, 12); b) considering (42) a pixel belonging to said image portion (20); c) determining (44, 46, 48) the absolute value of the difference between a light data item representative of the pixel in question and a light data item representative of a group of pixels (21), said group of pixels (21) comprising the pixel considered and at least two pixels belonging to different lines of said pixel considered; - d) incrementing (50, 54) by one step a value of a counter (24) when said absolute value is greater than a first predefined threshold (SA); e) repeating (52) steps b) to d) for each pixel of the image portion (20); and f) selecting (56) the frame encoding mode for the or each image portion (11, 12) when the counter value is greater than a second predefined threshold (SD) and the image encoding mode when the value counter is less than the second predefined threshold (SD).   2. Selection method according to claim 1, characterized in that said light data comprises a luminance data.   3. Selection process according to any one of claims 1 and 2, characterized in that the step of determining (44, 46, 48) the absolute value comprises a step of searching (46) the median value of the luminous data of the group of pixels (21). 2910216 14   4. Selection method according to any one of claims 1 to 3, characterized in that the image portion (20) comprises all the pixels of the image portion (11, 12) with the exception of pixels of the first I lines and the last n-1 lines, I being a natural integer.   5. A selection method according to any one of claims 1 to 4, characterized in that when the pixel in question belongs to the kth line, the group of pixels (21) comprises the pixel and the pixels between the k-1 line and the k + lth line.   6. Selection process according to any one of claims 4 and 5, characterized in that I is equal to 1.   A selection method according to any one of the preceding claims, characterized in that the image portion (11, 12) is a pair of macroblocks of said image, said pair of macroblocks comprising distributed pixels. according to 32 lines and 16 columns.   8. Selection method according to any one of the preceding claims, characterized in that it comprises, before step a) of defining an image portion (20), the following steps: - receive an image (3, 4); - consider a portion (11, 12) of the received image; and in that it comprises, after the selection step g), the following steps: - reset (62) the value of the counter (24) and consider another part (11, 12) of said image; and repeating steps a) to g) for all parts (11, 12) of the image. 5 10 30 2910216 15   9. Selection process according to any one of the preceding claims, characterized in that the image is an interlaced image.   A method of encoding at least one image (3, 4) according to the image mode / adaptive macro-block frame of the MPEG4 compression format, the method comprising the steps of the method of selecting an encoding mode according to any one of the preceding claims, and a step of encoding each portion of the image (11, 12) according to the selected encoding mode. 10   Recording medium comprising a computer program comprising instructions for executing the steps of the selection method according to any one of claims 1 to 9, when the program is executed on a computer. 15   12. A selection device (8) for a frame or frame coding mode of at least one part (11, 12) of an image (3, 4) of a video sequence (5), said part of images (11, 12) having pixels distributed along n rows and m columns, the selected encoding mode being for use in encoding the image portion (11, 12) according to the macroblock adaptive mode (MBAFF) of the MPEG4 compression format, characterized in that it comprises: determination means (17, 18, 19A, 19B) of the absolute value of the difference between a light datum of the pixel in question and a datum 25 light representative of a group of pixels (21), said group of pixels (21) comprising the pixel in question and at least two pixels belonging to different lines of said pixel in question; a counter (24) whose value is adapted to be incremented by one step when the absolute value calculated by the determination means (17, 18, 19A, 19B) is greater than a first predefined threshold (SA); and selection means (25, 26) adapted to select the frame coding mode for the or each picture portion (11, 12) when the counter value (24) is greater than a second predefined threshold ( SD), and the picture encoding mode when the value of the counter (24) is less than the second predefined threshold (SD).   13. A method of transcoding the MPEG2 compression format to the MPEG4 compression format, characterized in that the selection of a frame or frame coding mode for encoding at least a portion (11, 12) of an image a video sequence according to the macro-block adaptive mode (MBAFF) of the MPEG4 compression format, is performed according to the selection method of any one of claims 1 to 9.