KR100718468B1 - Method and device for video down-sampling transcoding - Google Patents

Abstract

The present invention relates to a method and apparatus for image reduction coding, and more particularly, to a method and apparatus for performing image reduction at high speed from H.264 to MPEG-4. An image reduction transcoding method for converting a first macroblock according to a first scheme into a second macroblock according to a second scheme includes: (a) a first condition in which the number of intra macroblocks included in the first macroblock is preset; Determining a block mode of the second macroblock, and converting to the determined block mode; (b) if the number of inter macroblocks included in the first macroblock satisfies the second condition, determining a block mode of the second macroblock and converting the block mode to the determined block mode; And (c) when the conversion of the macroblock is completed, performing motion vector readjustment on the second macroblock.

Image reduction, transcoding, macroblock, motion vector

Description

Method and device for video reduction transcoding {Method and device for video down-sampling transcoding}

1 is a diagram illustrating a method of reducing an image size in accordance with a preferred embodiment of the present invention.

2A to 2E are diagrams illustrating a macroblock mode and a method of converting a motion vector according to an exemplary embodiment of the present invention.

3 is a diagram illustrating a configuration of a transcoding apparatus for reducing image size according to a preferred embodiment of the present invention.

4 is a flowchart illustrating a method of reducing an image size according to a preferred embodiment of the present invention.

5A to 5D illustrate a method of reducing an image size according to a preferred embodiment of the present invention.

6A to 6D illustrate a macroblock mode and a method of converting a motion vector according to another preferred embodiment of the present invention.

<Description of the symbols for the main parts of the drawings>

300: transcoding device

310: first conversion unit

320: second conversion unit

330: vector readjustment unit

The present invention relates to a method and apparatus for image reduction coding, and more particularly, to a method and apparatus for performing image reduction at high speed from H.264 to MPEG-4.

Advances in computers and networks have made it possible to transmit a variety of data from the sender to the receiver. Among these technologies, multimedia standard compression formats are introduced to transmit a large amount of multimedia data through a network, and converted to each environment in consideration of QoS (Quality Of Service) of both sides in order to consider the environment of a transmitter and a receiver. Many video transcoding techniques have been introduced. Such video transcoding techniques are classified into transcoding methods in the pixel domain and transcoding methods in the DCT domain.

The best transcoding method in terms of image quality is a cascaded pixel-domain transcoding method of decoding a bitstream of an input image and re-coding the decoded image. However, this conventional technology has a problem that the complexity increases because the encoding is performed once more. Therefore, according to the prior art, there is no efficient transcoding method capable of improving the complexity while satisfying the image quality of the image.

Accordingly, an object of the present invention is to provide an image reduction transcoding method and apparatus capable of reducing complexity while satisfying an image quality by reusing macroblock information. That is, the present invention proposes a transcoding technique that can reuse information of each macro block (MB, Macro block) used in the H.264 decoder to reduce the complexity.

In addition, another object of the present invention is H.264 (ITU-T Recommendation H.264) and the Common Intermediate Format (CIF) encoded by ISO / IEC MPEG-4 Part 10 Advanced Video Coding (BP) baseline profile (BP) encoder. The video bitstream of size is reduced to half the size of the image in the pixel area and converted into MPEG-4 Simple profile (SP) video bitstream of QCIF (Quarter Common Intermediate Format) size for smooth communication. In order to provide a transcoding method and apparatus for carrying out the above, other objects of the present invention will become more apparent through the preferred embodiments described below.

According to the first aspect of the present invention to achieve the above object, it is possible to provide a method for performing image reduction at high speed from H.264 to MPEG-4.

According to a preferred embodiment of the present invention, in a video reduced transcoding method for converting a first macroblock according to a first scheme into a second macroblock according to a second scheme, (a) A (natural number) xB ( If the number of intra macroblocks included in the first macroblock of the size) satisfies a preset first condition, the block mode of the second macroblock of the size C (natural number) x D (natural number) is determined, and Converting to the determined block mode; (b) if the number of inter macroblocks included in the first macroblock satisfies a second condition, determining a block mode of the second macroblock and converting to the determined block mode; And (c) when the conversion of the first macroblock is completed, performing a motion vector readjustment on the second macroblock.

The image reduction transcoding method may be converted to the determined block mode while downsampling using an average filter.

The step (a) may include converting to one intra macroblock when the first macroblocks are all intra macroblocks.

The step (a) may include determining a block mode using a SAD comparison method of the macroblock when a part of the macroblock is an intra macroblock.

In the step (b), when the number of the inter macroblocks is equal to or greater than a preset number and the difference value between the motion vectors is calculated, when the difference value is less than or equal to the predetermined value, the mode is converted to inter mode. The method may include determining a block mode using a SAD comparison method of blocks and converting the block mode to the determined block mode.

The SAD comparison method may include calculating SADs of two preset inter blocks and determining a minimum inter SAD; And comparing the minimum inter SAD with an intra SAD of a preset intra block and converting the block to a block mode corresponding to the minimum SAD.

에 의하여 매크로블록의 화소값 평균이 산출되고-여기서, 상기 org_ij는 MB내에 있는 각각의 16x16 화소들의 값이고 상기 MB_mean은 MB내에 있는 화소값들의 평균을 지칭함-, 수학식

에 의해 상기 인트라 SAD가 산출되는 것을 특징으로 한다.The intra SAD is an equation

By means of which the pixel value average of the macroblock is calculated, wherein org _ij is the value of each 16x16 pixels in MB and MB _mean refers to the average of the pixel values in MB.

The intra SAD is calculated by.

에 의하여 수행되며, 여기서, 상기 m, n은 중간 단계인 MB이 취하는 8x8 블록들의 가로축, 세로축 좌표를 가리키고 상기 k, l은 2x2 MB들 안의 (m, n)번째 매크로블록에 대응하는 4x4 블록들의 가로축, 세로축 좌표를 지칭하는 것을 특징으로 한다.The motion vector readjustment is

Where m and n are the horizontal and vertical coordinates of the 8x8 blocks taken by the intermediate MB and k and l are the 4x4 blocks corresponding to the (m, n) th macroblocks in the 2x2 MBs. It is characterized by referring to the horizontal axis, vertical axis coordinates.

The step (c) may include: searching for an integer unit motion vector with respect to integer neighboring ± 3 neighboring pixels around the integer unit motion vector; And retrieving a motion vector of ± 1 neighboring pixels in a 1/2 pixel unit around the searched integer unit motion vector.

According to another preferred embodiment of the present invention, in a video reduced transcoding method for converting a first macroblock according to a first scheme into a second macroblock according to a second scheme, the first macroblock having a predetermined size Determining any one of the candidate block modes according to the second scheme as the block mode of the second macroblock determined by a condition set in advance by a type of macroblocks included in the macroblock, and converting the block mode to the determined block mode; And when the conversion of the first macroblock is completed, performing a motion vector readjustment on the second macroblock.

The preset condition may be to determine a block mode with one intra macroblock according to the second scheme when three or more intra macroblocks are included in the first macroblock.

The preset condition is the candidate block mode using a sum of absolute difference (SAD) comparison method of each macroblock included in the first macroblock when two or less intra macroblocks are included in the first macroblock. It may be determined by either. Here, the candidate block mode may include at least one of a skip (SKIP) mode, an inter 16x16 mode, and an inter 8x8 mode.

The preset condition is a SAD (Sum of Absolute Difference) comparison scheme of each inter macroblock when the first macroblock includes only inter macroblocks and includes two or less inter 16x16 macroblocks among the inter macroblocks. It may be determined by using any one of the candidate block mode using. Here, the candidate block mode may include at least one of a skip (SKIP) mode, an inter 16x16 mode, and an inter 8x8 mode.

After calculating the difference value between the motion vectors of each macroblock, if the difference value is less than or equal to a preset threshold, one of the inter 16x16 mode and the skip mode may be determined. If not, the inter 8x8 mode may be determined. .

One of the candidate block modes determined is a block mode having a minimum SAD value by the SAD comparison method.

에 의하여 매크로블록의 화소값 평균이 산출되고-여기서, 상기 org_ij는 MB내에 있는 각각의 16x16 화소들의 값이고 상기 MB_mean은 MB내에 있는 화소값들의 평균을 지칭함-, 수학식

에 의해 상기 인트라 매크로블록의 SAD가 산출되는 것을 특징으로 한다.SAD of the intra macroblock is,

By means of which the pixel value average of the macroblock is calculated, wherein org _ij is the value of each 16x16 pixels in MB and MB _mean refers to the average of the pixel values in MB.

It is characterized in that the SAD of the intra macroblock is calculated by.

를 이용하 여 수행되며, 상기 m, n은 중간 단계인 MB이 취하는 8x8 블록들의 가로축, 세로축 좌표를 가리키고 상기 k, l은 2x2 MB들 안의 (m, n)번째 매크로블록에 대응하는 4x4 블록들의 가로축, 세로축 좌표를 지칭하는 것을 특징으로 한다.The motion vector readjustment is

M and n denote the horizontal and vertical coordinates of the 8x8 blocks taken by the intermediate MB, and k and l denote the horizontal axes of the 4x4 blocks corresponding to the (m, n) th macroblocks in the 2x2 MBs. , It refers to the vertical axis coordinates.

In accordance with a second aspect of the present invention, an apparatus for performing image reduction at high speed from H.264 to MPEG-4 can be provided.

According to a preferred embodiment of the present invention, an image reduction transcoding apparatus for converting a first macroblock according to a first scheme into a second macroblock according to a second scheme, wherein the intra included in the first macroblock A first converter configured to determine a block mode of the second macroblock and to convert to the determined block mode when the number of macroblocks meets a first predetermined condition; A second converter configured to determine a block mode of the second macroblock and convert to the determined block mode when the number of inter macroblocks included in the first macroblock satisfies a second condition; And a motion vector readjustment unit configured to perform a motion vector readjustment for the second macroblock when the first macroblock is converted.

If the first macroblocks are all intra macroblocks, the first conversion unit converts one intra macroblock to a single macroblock, and if a part of the macroblocks is an intra macroblock, the first macroblock is a block using an SAD comparison method of the macroblocks. The mode can be determined.

The second converting unit calculates a difference value between the motion vectors when the number of the inter macroblocks is equal to or greater than a preset number, and when the difference value is smaller than or equal to the preset value, converts to the inter mode. The block mode may be determined by using the SAD comparison method, and the block mode may be converted into the determined block mode.

The SAD comparison method calculates a SAD of two preset inter blocks, determines a minimum inter SAD, compares the minimum inter SAD with an intra SAD of a preset intra block, and converts the block mode corresponding to the minimum SAD. Characterized in that.

According to another preferred embodiment of the present invention, an image reduction transcoding apparatus for converting a first macroblock according to a first scheme into a second macroblock according to a second scheme, wherein the first macroblock having a predetermined size A converter that determines one of the candidate block modes according to the second scheme determined by a condition set in advance by a type of macroblocks included in the block mode of the second macroblock, and converts the block mode to the determined block mode. ; And a motion vector readjustment unit configured to perform a motion vector readjustment for the second macroblock when the first macroblock is converted.

The preset condition may be to determine a block mode with one intra macroblock according to the second scheme when three or more intra macroblocks are included in the first macroblock.

The preset condition is the candidate block mode using a sum of absolute difference (SAD) comparison method of each macroblock included in the first macroblock when two or less intra macroblocks are included in the first macroblock. It may be determined by either. Here, the candidate block mode may include at least one of a skip (SKIP) mode, an inter 16x16 mode, and an inter 8x8 mode.

The preset condition is a SAD (Sum of Absolute Difference) comparison scheme of each inter macroblock when the first macroblock includes only inter macroblocks and includes two or less inter 16x16 macroblocks among the inter macroblocks. It may be determined by using any one of the candidate block mode using. Here, the candidate block mode may include at least one of a skip (SKIP) mode, an inter 16x16 mode, and an inter 8x8 mode.

After calculating the difference value between the motion vectors of each macroblock, the converter determines one of the inter 16x16 mode and the skip mode when the difference value is less than or equal to a preset threshold. Otherwise, the converter determines the inter 8x8 mode. It is characterized by determining.

에 의하여 매크로블록의 화소값 평균이 산출되고-여기서, 상기 org_ij는 MB내에 있는 각각의 16x16 화소들의 값이고 상기 MB_mean은 MB내에 있는 화소값들의 평균을 지칭함-, 수학식

에 의해 상기 인트라 매크로블록의 SAD가 산출되는 것을 특징으로 한다.SAD of the intra macroblock is,

By means of which the pixel value average of the macroblock is calculated, wherein org _ij is the value of each 16x16 pixels in MB and MB _mean refers to the average of the pixel values in MB.

It is characterized in that the SAD of the intra macroblock is calculated by.

를 이용하여 수행되며, 상기 m, n은 중간 단계인 MB이 취하는 8x8 블록들의 가로축, 세로축 좌표를 가리키고 상기 k, l은 2x2 MB들 안의 (m, n)번째 매크로블록에 대응하는 4x4 블록들의 가로축, 세로축 좌표를 지칭하는 것을 특징으로 한다.The motion vector readjustment is

M and n denote horizontal and vertical coordinates of the 8x8 blocks taken by the intermediate MB, and k and l denote the horizontal axes of the 4x4 blocks corresponding to the (m, n) th macroblocks in the 2x2 MBs. , It refers to the vertical axis coordinates.

Hereinafter, exemplary embodiments of an image reduction transcoding method and apparatus according to the present invention will be described in detail with reference to the accompanying drawings. In the following description with reference to the accompanying drawings, the same or corresponding components may be used. The same reference numerals will be given, and redundant description thereof will be omitted.

In general, the simplest method of converting MBs (macroblocks) in H.264 frames to MBs in MPEG-4 frames is to restore all compressed H.264 frames as input, and then to restore the MPEG-4 video. It is a serial pixel region transcoding method that compresses again. However, since the motion prediction process for all MBs in all frames is performed by the MPEG-4 encoder, the method may cause a problem in real time transmission in a transcoder due to a large amount of computation.

In order to solve this problem, the present invention is configured to reuse the block mode and motion vector information used in the H.264 decoder in the MPEG-4 encoder in order to be suitable for real time transmission without performing the motion prediction process. According to an embodiment of the present invention, information on a frame format, MB information, and motion vector currently reconstructed in the process of performing variable length decoding (VLD) on a H.264 compressed bitstream Figure out. Here, if the frame format of the reconstructed bitstream is a P frame (Predictive frame), the motion prediction process that consumes much time in the MPEG-4 encoder can be omitted by reusing the block type and the motion vector in the MB.

As described above, the transcoding apparatus according to the present invention may include an H.264 decoder and an MPEG4 encoder, and the information identified in the H.264 decoder may be provided as an MPEG4 encoder. The transcoding device according to the present invention may be provided in a media gateway server for transmitting a bitstream to a receiving device.

Hereinafter, the transcoding method performed in the transcoding apparatus according to the present invention will be described. Those skilled in the art will be able to easily understand the configuration and operation of the transcoding apparatus according to the present invention through the following description.

Hereinafter, referring to FIGS. 1 and 2A to 2E, transcoding using block vector conversion and motion vector adjustment in the pixel domain when converting from H.264 baseline profile (MPP) to MPEG-4 SP (simple profile) will be described. The coding method is explained. A configuration of the transcoding apparatus will be described with reference to FIG. 3, and a specific transcoding method will be described with reference to FIG. 4. In addition, with reference to Figures 5a to 5d will be described the experimental results when the transcoding H.264 BP by the pixel region transcoding method to MPEG-4 SP. Finally, with reference to FIGS. 6A to 6D, a method of transcoding using block-to-block type conversion and motion vector adjustment by a statistical method will be described.

Procedure of Image Reduction Transcoding

FIG. 1 illustrates a method for downsizing an image using an average filter as an image reduction method according to a preferred embodiment of the present invention.

For example, in case of converting 4 MB types into 1 MB type, according to the prior art, while converting a CIF sized image passed through an H.264 (SP) encoder into a QCIF sized image again, The source image is deteriorated in image quality through two encoding processes (encoding and down-sampling).

According to an embodiment of the present invention, four 16x16 MB (macro blocks) in the CIF image reconstructed by the H.264 decoder are reduced to one 16x16 MB to be encoded by the MPEG-4 encoder and converted into a QCIF image. According to the present invention, in order to reduce a CIF image to a QCIF image, a method of sampling the one pixel value by average filtering the four pixel values, and taking only one pixel value among the four pixel values, Among sub-sampling methods, a method of reducing an image using an average filter is used. This is because the sub-sampling technique reduces the image and provides some gains in terms of time complexity. However, the average filter is used because the deterioration of image quality is remarkable. In the present invention, a transcoding method in a pixel domain is used without suggesting a transcoding method in a DCT (Discrete Cosine Transform) region having low computational complexity. This is because of H.264 non-linear loop filter, DCT the loop filtered image again in order to transcode in the DCT domain, and then transcode in the DCT domain, between 4x4 integer DCT and 8x8 DCT. This is because a large amount of calculation is required because a conversion equation must be derived.

In H.264, 1/4 pixel ME (Motion Estimation) and MC (Motion Compensation), 7 variable blocks (i.e. Inter 16x16, Inter 16x8, Inter 8x16) , Inter 8x8, Inter 8x4, Inter 4x8, and Inter 4x4) and 10 block modes (i.e. Skip, Intra 16x16, Intra 4x4, Inter (4x4) Inter 16x16, Inter 16x8, Inter 8x16, Inter 8x8, Inter 8x4, Inter 4x8 and Inter 4x4 modes), but MPEG-4 is intra Because it supports (Intra) 16x16, Inter 16x16, Inter 8x8, and skip mode, the present invention is configured to perform block shape conversion to reuse such information in MPEG-4.

2A to 2E illustrate a method of converting an H.264 four MB mode and a motion vector into an MPEG-4 MB mode and a motion vector according to a preferred embodiment of the present invention.

Hereinafter, referring to various embodiments of converting the format (MB mode) of 2x2 MBs of the reconstructed H.264 bitstream into one MB form that can be reused in MPEG-4, MPEG-4 SP in H.264 BP The following describes a transcoding method using a transformation between block types and a motion vector adjustment in a pixel domain when converting to.

All Skip modes in H.264 according to FIG. 2A represent a preprocessing step for converting 2x2 MBs in H.264 into a single MB in MPEG-4, and in Inter16x16 mode in H.264 as shown in FIG. 2A. Convert

If 2x2 MBs of H.264 are composed only of MBs that are Intra 16x16 or Intra4x4, as shown in FIG. 2B, convert to one Intra 16x16 MB and use an average filter. Reduce the image.

When only two or three MBs of 2x2 MBs of H.264 are configured as Intra mode MBs as shown in FIG. 2C, Inter 16x16, Inter8x8 and Intra (x) in the MPEG-4 encoder are used. Intra) It is unclear whether to convert to 16x16 mode.

In this case, according to the prior art, when the conversion to a specific mode (Inter 16x16, Inter 8x8 or Intra 16x16 mode) in MPEG-4, the image quality is reduced and the amount of encoded bits An increasing phenomenon appears.

Therefore, in the present invention, the sum of SAD (Sum of Absolute Difference) value (SAD-128) of four Inter 8x8 blocks and an SAD value of four Inter 8x8 blocks in an MPEG-4 encoder is shown. Compare with and select the mode with the lowest SAD value. Subsequently, the selected inter mode SAD value is compared with the intra mode SAD value calculated by Equations 1 and 2 below to finally select a mode having a minimum SAD value.

Where org _ij is the value of each 16x16 pixel in MB and MB _mean is the average of the pixel values in MB. By converting to the optimal mode selected in this manner, it is possible to satisfy both the image quality aspect and the bit amount encoded.

As shown in FIG. 2D, when there are four or three MBs having an Inter 16x16 mode in 2x2 MBs of H.264, the mobile terminal is configured to convert to the Inter16x16 mode in the MPEG-4 encoder. According to the prior art, if there are four or three MBs having an Inter16x16 mode in 2x2 MBs of H.264, they are converted to Inter 16x16 or Inter 8x8 mode in the MPEG-4 encoder. Although this is possible, the present invention converts the Inter 16x16 mode instead of the Inter 8x8 mode. First, the average value of each quarter-unit motion vector of each MB in a 2x2 MB is taken as an integer unit motion vector of four 8x8 blocks in MPEG-4 MB, and then one motion is used as the average value of four 8x8 block motion vectors thus obtained. Mode conversion is achieved by taking the vector at Inter 16x16 MB. A specific calculation equation for obtaining the integer unit motion vector is shown in Equation 3 below.

Here, m and n refer to the horizontal and vertical coordinates of the 8x8 blocks taken by the intermediate MB, and k and l refer to the horizontal and vertical coordinates of the 4x4 blocks taken by the (m, n) th MB in the 2x2 MBs.

The reason for converting to Inter16x16 mode instead of Inter8x8 mode is to fix the QP (quantization parameter) to 10 for foreman video when transcoding from H.264 to MPEG-4. In the same condition as in FIG. 2D, the PSNR is improved by 0.08dB from 31.61dB to 31.69dB when the Inter 8x8 mode is converted to the Inter 8x8 mode. Because the coded bit amount increases from 143.2kbps to 156.8kbps, the overall coding efficiency decreases.

Referring to FIG. 2E, H.264 2x2 MBs are complex block modes, that is, MBs in one or more InterP8x8 modes, one or less Intra mode MBs and other Inter mode MBs. In this case, each MB in 2x2 MBs is converted into an 8x8 block in MPEG-4. The integer motion vector taken by an 8x8 block in MPEG-4 is obtained by calculating the average value of the quarter-unit motion vector of 16 4x4 blocks and dividing the value by four again to obtain an integer motion vector. The calculation formula is the same as the above equation (3).

3 is a diagram showing the configuration of a transcoding device according to a preferred embodiment of the present invention.

Referring to FIG. 3, the transcoding apparatus 300 according to the present invention includes a first transform unit 310, a second transform unit 320, and a motion vector readjustment unit 330.

When the number of intra macroblocks included in the first macroblock satisfies a first predetermined condition, the first converter 310 determines a block mode of the second macroblock and converts the block mode to the determined block mode. . For example, when the first macroblocks are all intra macroblocks, the first converter 310 converts the data into one intra macroblock (see FIG. 2B), and when a part of the macroblocks is an intra macroblock, the macro The block mode is determined using the SAD comparison method of the blocks (see FIG. 2C).

When the number of inter macroblocks included in the first macroblock satisfies the second condition, the second converter 320 determines a block mode of the second macroblock and converts the block mode to the determined block mode. For example, when the number of the inter macroblocks is equal to or greater than the preset number and the difference value between the motion vectors, the second transforming unit 320 converts to the inter mode when the difference is less than or equal to the preset value. In other cases, the block mode is determined by using the SAD comparison method of the macroblock, and the block mode is converted to the determined block mode. More detailed functions will be described with reference to the flowchart of FIG. 4.

The motion vector readjustment unit 330 performs motion vector readjustment for the second macroblock.

FIG. 4 shows a specific flow chart of transcoding from MPEG-2 to H.264 based on 2x2 MB.

As shown in FIG. 4, when the H.264 2x2 MB form is input (step S410), the input H.264 2x2 MB form is converted into MPEG-4 1x1 MB form according to each condition (step S420). Then, down sampling is performed using the average filter (step S430).

Then, the mode decision for the MPEG4 block type is performed.

If the number of intra macroblocks in the 2x2 MBs of H.264 is greater than two in step S425, the flow advances to step S430 to determine whether the number of macroblocks is four, and if not four, step S435 The process proceeds to Intra mode, and the process proceeds to step S455 in other cases.

However, if the number of intra macroblocks is within 2x2 MBs of H.264 in step S425, the flow proceeds to step S440. If there are more than two Inter 16x16 modes (for example, four or three) in 2x2 MBs of H.264 in step S440, the flow proceeds to step S445 to convert to Inter 16x16. . However, in step S440, if there are two or less Inter16x16 modes in 2x2 MBs of H.264, the difference between all integer motion vectors in step S450 is compared with a preset threshold (for example, 4). .

If the difference is greater than or equal to the threshold, the process proceeds from step S450 to step S455 to determine the MB mode through the SAD process. However, if the difference is less than the preset threshold, the process proceeds to step 445 to convert to inter 16x16.

Here, in the H.264 encoder, the mode determined by the edge such as Inter8x8, Inter8x4, Inter4x8, and Inter4x4 modes is used. The reason for converting to Inter16x16 MB mode instead of 8x8 MB mode is that H.264 encoding and image reduction are a kind of low pass filtering (LPF).

Thereafter, the mode determination of the macroblock is terminated in step S460, and the flow proceeds to step S415 to start conversion of the next macroblock.

When the mode decision is completed by the above-described process, the motion vector readjustment is performed in step S470. That is, the integer unit motion vector is searched for the neighboring integer unit ± 3 neighbor pixels around the calculated integer unit motion vector, and ± 1 unit of 1/2 pixel unit around the found integer unit motion vector is found. By finding a motion vector for a neighboring pixel, a motion vector refinement may be performed to obtain PSNR (Peak Signal-to-Noise Ratios).

5A to 5D are diagrams illustrating an experimental result when H.264 BP according to a preferred embodiment of the present invention is transcoded by pixel region transcoding using MPEG-4 SP.

The experiment was carried out using a transcoder implemented using a JM (Joint Model) decoder reconstructed with H.264 BP and a MoMuSys -FDIS-V1.0 encoder reconstructed as an SP of MPEG-4. It was. Experiments were performed on a Pentium IV 2.8GHz PC, and H.264 bitstreams generated by compressing Foreman, News, Paris and Coast CIF (352x288) images in a JM encoder were used. For each image used in the bitstream generation, only the first frame was used an I (Intra) frame, the rest were all P (Predictive) frames, and the experiment was performed by compressing 300 sequences at a frame rate of 30 Hz. Was

For the Coast of FIG. 5A, the Paris of FIG. 5B, the News of FIG. 5C and the Foreman sequence of FIG. 5D, the serial pixel region transcoding schemes 510, 520, 530, 540, the PSNR of the brightness (Y) signal between the image compression methods 515, 525, 535, 545 according to the conversion method according to the present invention is shown. Here, almost no image degradation was observed in all images except the foreman image of FIG. 5D having a scene change, and the image quality of up to 0.5 dB was improved in the coast image of FIG. 5A.

Table 1 below shows the temporal gains according to the transcoding scheme according to the present invention. Although there is a slight difference in each image, the average speed is improved by 1.72 times, and it can be seen that the speed is improved by about 4.1 times considering only the encoder (MPEG-4 encoder).

Sequence Cascade pixel-domain transcoding Proposed Transcoding H.264 decoder time MPEG-4 encoder time Transcoder Total time H.264 decoder time MPEG-4 encoder time Transcoder Total time News 34.14 52.23 86.37 34.14 12.1 46.24 Paris 37.2 52.18 89.38 37.2 12.3 49.5 Coast 51.48 53.54 105.02 51.48 13.29 64.77 Foreman 49.54 53.49 103.03 49.54 13.32 62.86

As described above, according to the present invention, an information conversion method may be provided to use block information defined in H.264 again in MPEG-4. According to the prior art, it can be seen that there is almost no deterioration in image quality while showing 1.8 times the performance improvement by reusing the previous motion vector and the block information in H.264 compared to the serial pixel region transcoding method which has a high cost in time.

So far, the method of converting from H.264 baseline profile (BP) to MPEG4 SP (simple profile) according to the present invention has been described. In addition, when converting the block mode defined in H.264 to the block mode defined in MPEG4, the description has been focused on a method that considers all the block modes defined in MPEG4.

However, when statistically reviewed, when converting 2x2 MB of H.264 to 1x1 MB of MPEG4, considering the characteristics of 2x2 MB of H.264, only the minimum MPEG4 block mode can be considered, and there is almost no deterioration in image quality. We found that transcoding is possible quickly.

Table 2 below shows the conversion frequency of MPEG-4 block mode according to the characteristics of 2x2 MB of H.264 statistically examined, and FIGS. 6A to 6D illustrate the conversion process of the macroblock mode and the motion vector considering the statistical method. Is illustrated.

2 MB modes of the H.264MB modes of the MPEG-4 2

2 MB modes of the H.264 INTRA SKIP INTER 16 ㅧ 16 INTER 8 ㅧ 8 Total Remarks two Inter16

16, two blocks smaller than Inter16

8 or Inter8

16 0.20% 13.53% 3.48% 1.88% 19.09% See FIG. 6A one Inter16

16, three blocks smaller than Inter16

8 or Inter8

16 0.22% 13.91% 2.16% 3.21% 19.49% four blocks smaller than Inter16

8 or Inter8

16 0.14% 8.55% 0.86% 3.25% 12.79% four Inter16

16 0.26% 3.48% 25.94% 0.09% 29.77% See Figure 6b three Inter16

16, one block smaller than Inter16

8 or Inter8

16 0.13% 8.01% 5.75% 0.47% 14.35% one Intra, three Inter Fig. 5 (c) 0.03% 1.15% 0.32% 1.26% 2.76% See Figure 6c two Intra, two Inter Fig. 5 (c) 0.02% 0.37% 0.10% 0.65% 1.14% three Intra, one Inter Fig. 5 (d) 0.00% 0.10% 0.04% 0.16% 0.31% See Figure 6d four Intra Fig. 5 (d) 0.05% 0.09% 0.01% 0.14% 0.29% Total Number 1.04% 49.19% 38.65% 11.11% 100.00%

As shown in Table 2 and FIG. 6, the transcoding method according to the present invention may consider only the block mode of MPEG4 having a relatively high frequency when converting a MB of a predetermined size from H.264 to MPEG4. At this time, a characteristic of 2x2 MB of H.264 may be used.

For example, if two macroblocks of 2x2 MB of H.264 are inter16x16 macroblocks, and the remaining two macroblocks are smaller than inter16x8 or inter8x16 macroblocks, the inter16x16 block mode of MPEG4 is illustrated in FIG. 6A. Only the skip mode or the inter 8x8 block mode may be considered in the conversion. It can be seen that the corresponding block modes have a relatively high frequency when referring to Table 2. In addition, as shown, whether to select one of the inter 16x16 block mode or the skip mode of MPEG4 or the inter 8x8 block mode, each difference value (difference) of the motion vector in the unit of 1/4 pixel is a threshold value (for example, 9). May be determined primarily by

However, when three or more macroblocks of 2x2 MB of H.264 are intra macroblocks, the block mode of MPEG4 having the highest conversion frequency in Table 2 is an inter 8x8 block mode. However, since the frequency shows a very low value (i.e., 0.14% or 0.16%), it can be ignored and the block mode of MPEG4 can be determined as the intra block mode as shown in FIG. 6D.

As described above, the present invention can provide an image reduction transcoding method and apparatus capable of reducing complexity while satisfying an image quality by reusing information of a macroblock.

In addition, the transcoder for converting from H.264 to MPEG-4 according to the present invention can enable the use of H.264 compressed video, which is a new video compression standard, without replacing equipment that used the conventional MPEG-4 codec. There is also an effect.

Although the above has been described with reference to a preferred embodiment of the present invention, those skilled in the art to which the present invention pertains without departing from the spirit and scope of the invention as set forth in the claims below It will be appreciated that modifications and variations can be made.

Claims (28)

An image reduction transcoding method for converting a first macroblock according to a first scheme into a second macroblock according to a second scheme, (a) if the number of intra macroblocks included in the first macroblock of A (natural number) x B (natural number) satisfies a preset first condition, the second macro of C (natural number) x D (natural number) size Determining a block mode of the block and converting to the determined block mode; (b) if the number of inter macroblocks included in the first macroblock satisfies a second condition, determining a block mode of the second macroblock and converting to the determined block mode; And (c) if the conversion of the first macroblock is completed, performing motion vector readjustment on the second macroblock; In the step (a), when the first macroblocks are all intra macroblocks, the macroblock is converted into one intra macroblock, and when some of the macroblocks included in the first macroblock are intra macroblocks, each macroblock The block mode is determined using the SAD comparison method of, In the step (b), when the number of the inter macroblocks is equal to or greater than a preset number and a difference value between motion vectors is calculated, when the difference value is less than or equal to the preset value, the inter mode is converted into an inter mode. 1 determines a block mode by using the SAD comparison method of the macroblock and converts it to the determined block mode, The SAD comparison method determines a minimum inter SAD by calculating SADs of two preset inter blocks, and then converts the minimum inter SADs into a block mode corresponding to the minimum SAD by comparing the minimum SADs with a preset intra block. An image reduction transcoding method, characterized in that. The method of claim 1, And converting to the determined block mode while downsampling using an average filter. delete delete delete delete The method of claim 1, The intra SAD is Equation

By means of which the pixel value average of the macroblock is calculated, wherein org _ij is the value of each 16x16 pixels in MB and MB _mean refers to the average of pixel values in MB; Equation

The intra SAD is calculated by the image reduction transcoding method. The method of claim 1, The motion vector readjustment is

Where m and n are the horizontal and vertical coordinates of the 8x8 blocks taken by the intermediate MB and k and l are the 4x4 blocks corresponding to the (m, n) th macroblocks in the 2x2 MBs. An image reduction transcoding method characterized by referring to a horizontal axis and a vertical axis coordinate. The method of claim 1, Step (c) is Retrieving an integer unit motion vector with respect to an integer unit ± 3 neighboring pixels around the integer unit motion vector; And And retrieving a motion vector of ± 1 neighboring pixels in a unit of 1/2 pixel around the retrieved integer unit motion vector. An image reduction transcoding method for converting a first macroblock according to a first scheme into a second macroblock according to a second scheme, Determine one of the candidate block modes according to the second scheme as the block mode of the second macroblock, which is determined by a condition preset by a type of macroblocks included in the first macroblock of a predetermined size, Converting to the determined block mode; And When the conversion of the first macroblock is completed, performing a motion vector readjustment for the second macroblock, The preset condition may be skipped using a sum of absolute difference (SAD) comparison method of each macroblock included in the first macroblock when two or less intra macroblocks are included in the first macroblock. Determine any one of the candidate block modes including one or more of a mode, an Inter 16x16 mode, and an Inter 8x8 mode, The SAD comparison method determines a minimum inter SAD by calculating SADs of two preset inter blocks, and then converts the minimum inter SADs into a block mode corresponding to the minimum SAD by comparing the minimum SADs with a preset intra block. Image reduction transcoding method, characterized in that. The method of claim 10, The preset condition is a video reduction transcoding method characterized in that the block mode is determined by one intra macroblock according to the second method when three or more intra macroblocks are included in the first macroblock. delete The method of claim 10, The preset condition is that if the first macroblock is composed of only inter macroblocks and two or less inter 16x16 macroblocks among the inter macroblocks are included, the candidate block using the SAD comparison method of each inter macroblock. An image reduction transcoding method, characterized in that determined by any one of the modes. The method according to claim 10 or 13, After calculating the difference value between the motion vectors of each macroblock, if the difference value is less than or equal to a predetermined threshold value, one of the inter 16x16 mode and the skip mode is determined, but otherwise, the inter 8x8 mode is determined. An image reduction transcoding method. delete The method of claim 10, SAD of the intra macroblock, Equation

By means of which the pixel value average of the macroblock is calculated, wherein org _ij is the value of each 16x16 pixels in MB and MB _mean refers to the average of pixel values in MB; Equation

SAD of the intra macroblock is calculated by using the image reduction transcoding method. The method of claim 10, The motion vector readjustment is

M and n denote horizontal and vertical coordinates of the 8x8 blocks taken by the intermediate MB, and k and l denote the horizontal axes of the 4x4 blocks corresponding to the (m, n) th macroblocks in the 2x2 MBs. And transversal transcoding method according to the vertical axis. An image reduction transcoding apparatus for converting a first macroblock according to a first scheme into a second macroblock according to a second scheme, A first converter determining a block mode of the second macroblock and converting the block mode to the determined block mode when the number of intra macroblocks included in the first macroblock satisfies a preset first condition; A second converter configured to determine a block mode of the second macroblock and convert to the determined block mode when the number of inter macroblocks included in the first macroblock satisfies a second condition; And Comprising a motion vector readjustment unit for performing a motion vector readjustment for the second macroblock when the conversion of the first macroblock is completed, The first converter converts one intra macroblock when all of the first macroblocks are intra macroblocks, and SAD of each macroblock when some of the macroblocks included in the first macroblock are intra macroblocks. The block mode is determined using a comparison method, The second converter calculates a difference value between the motion vectors when the number of the inter macroblocks is equal to or larger than a preset number, and converts the inter macroblock to an inter mode when the difference value is smaller than or equal to a preset value. Determining the block mode using the SAD comparison method of the block to convert to the determined block mode, The SAD comparison method determines a minimum inter SAD by calculating SADs of two preset inter blocks, and then converts the minimum inter SADs into a block mode corresponding to the minimum SAD by comparing the minimum SADs with a preset intra block. An image reduction transcoding device, characterized in that. delete delete delete An image reduction transcoding apparatus for converting a first macroblock according to a first scheme into a second macroblock according to a second scheme, Determine one of the candidate block modes according to the second scheme as the block mode of the second macroblock, which is determined by a condition preset by a type of macroblocks included in the first macroblock of a predetermined size, A conversion unit for converting to the determined block mode; And Comprising a motion vector readjustment unit for performing a motion vector readjustment for the second macroblock when the conversion of the first macroblock is completed, The preset condition may be skipped using a sum of absolute difference (SAD) comparison method of each macroblock included in the first macroblock when two or less intra macroblocks are included in the first macroblock. Determine any one of the candidate block modes including one or more of a mode, an Inter 16x16 mode, and an Inter 8x8 mode, The SAD comparison method determines a minimum inter SAD by calculating SADs of two preset inter blocks, and then converts the minimum inter SADs into a block mode corresponding to the minimum SAD by comparing the minimum SADs with a preset intra block. An image reduction transcoding device, characterized in that. The method of claim 22, The preset condition is an image reduction transcoding device, characterized in that the block mode is determined by one intra macroblock according to the second method when three or more intra macroblocks are included in the first macroblock. delete The method of claim 22, The preset condition is that if the first macroblock is composed of only inter macroblocks and two or less inter 16x16 macroblocks among the inter macroblocks are included, the candidate block using the SAD comparison method of each inter macroblock. An image reduction transcoding device, characterized in that determined by any one of the modes. The method of claim 22 or 25, After calculating the difference value between the motion vectors of each macroblock, the converter determines one of the inter 16x16 mode and the skip mode when the difference value is less than or equal to a preset threshold. And a video reduction transcoding device. The method of claim 22, SAD of the intra macroblock, Equation

By means of which the pixel value average of the macroblock is calculated, wherein org _ij is the value of each 16x16 pixels in MB and MB _mean refers to the average of pixel values in MB; Equation

The SAD of the intra macroblock is calculated by the apparatus. The method of claim 22, The motion vector readjustment is

M and n denote horizontal and vertical coordinates of the 8x8 blocks taken by the intermediate MB, and k and l denote the horizontal axes of the 4x4 blocks corresponding to the (m, n) th macroblocks in the 2x2 MBs. And a vertical axis coordinate.

Images