KR20100032235A - Method and apparatus for transcoding compressed image - Google Patents

Abstract

PURPOSE: A method and apparatus for compressed video image transcoding is provided to convert compressed image to MPEG-4(Moving Picture Experts Group-4) SP(Simple Profile) format. CONSTITUTION: An MVC(decoder)(10) decodes bitstream which is compressed by a certain first compression; The MVC decoder output data of pixel region, motion information of each macroblock and macroblock information. A MPEG-4[encoder](20) receives data of pixel region. The MPEG-4 encoder performs motion compensation using motion vector information and macroblock information.

Description

Compressed image transcoding method and apparatus {Method and apparatus for transcoding compressed image}

The present invention relates to a transcoding method and apparatus for converting an image compressed in a predetermined format into another image compression format. More specifically, the present invention relates to MPEG-, which has been widely used in the past, for compressing an image compressed by an MVC (Multi-view Video Coding) standard for efficiently compressing an image obtained by using two or more cameras for an object. 4 Method and apparatus for converting into Simple profile (SP) format.

The present invention is the result of a research project (video compression technology and copyright management technology research for the next generation DMB / DTV) included in the Seoul-Academic-Industrial Cooperation Project (New Technology R & D Support Project).

As computers and communication technologies have evolved, a variety of data can be transmitted over the network. In order to transmit a large amount of multimedia data through a network, various multimedia data compression formats have been introduced. Since the environment of the sender and the receiver is different, one data format is considered in consideration of the quality of service (QoS) of both parties. It is necessary to convert to a format, which is called transcoding. Video transcoding includes a transcoding method in a pixel-domain and a transcoding method in a frequency conversion region, for example, a discrete cosine transform (DCT) region.

To perform transcoding in the pixel domain, decoding and encoding must be performed again. Therefore, motion prediction for all macroblocks in all frames must be performed again. Therefore, transcoding takes a long time and it is difficult to transmit the transcoded image in real time.

According to the present invention, when decoding and encoding have to be performed again in order to perform transcoding in the pixel region, transcoding takes a long time and transcodes because the motion prediction process for all macroblocks in all frames has to be performed again. In order to overcome the difficulty of transmitting images in real time, an object of the present invention is to provide a method and apparatus for converting an image compressed by the MVC (Multi-view Video Coding) standard into MPEG-4 Simple profile (SP) format. .

In order to achieve the above object, the present invention comprises the steps of: (a) decoding the bitstream compressed by a predetermined first compression method to output data of the pixel region, motion vector information and macroblock information of each macroblock; And (b) receiving data of the pixel region and performing encoding using a predetermined second compression method, wherein the motion compensation is performed using the motion vector information and macroblock information. do.

In the step (b), the block mode information and the motion vector information of the macroblock constituting the image produced by the first compression method are received, the block mode is converted into a 16x16 or 8x8 block mode, and the motion It is preferable to map the vector to a motion vector of a 16x16 or 8x8 block.

In addition, the present invention to achieve the above object (a) receiving the block mode information and motion vector information of the macroblock constituting the image produced by a predetermined compression method; (b) converting the block mode to a 16x16 or 8x8 block mode; And (c) mapping the motion vector to a motion vector of a 16x16 or 8x8 block.

In the transcoding method, (d) a half pixel is formed around a pixel of a motion vector coordinate obtained by performing a motion estimation on a pixel neighboring the pixel of the mapped motion vector coordinate, The method may further include determining a motion vector by performing motion estimation on the motion vector.

On the other hand, according to another aspect of the present invention for achieving the above object, a decoder for decoding the bitstream compressed by a predetermined first compression method to output the data of the pixel region, the motion vector information and macroblock information of each macroblock ; And an encoder configured to perform motion compensation using the motion vector information and the macroblock information when the data of the pixel area is input and encoded by a predetermined second compression method.

The encoder includes a motion compensator, wherein the motion compensator receives the block mode information and the motion vector information of the macroblock constituting the image produced by the first compression method, and converts the block mode to 16x16 or 8x8 block mode. Preferably, the motion vector is mapped to a motion vector of a 16x16 or 8x8 block.

According to the present invention, a video compressed according to the MVC standard can be played in an MPEG-4 terminal. In other words, it is possible to play back video created according to the new standard MVC standard without replacing the conventional MPEG-4 terminal to prevent waste of resources.

Hereinafter, exemplary embodiments of the present invention will be described in detail with reference to the accompanying drawings.

The present invention relates to MVC (Multi-view Video Coding, H.264 / MPEG-4 Part 10 AVC Amendment 4: Multi-view Video), a standard for efficiently compressing an image obtained by using two or more cameras for an object. Disclosed is a transcoding method and apparatus for converting an image compressed by Coding) into an MPEG-4 Simple profile (SP) format which is widely used in the related art.

MVC is a compression standard that enables H.264-based compression of images obtained by using two or more cameras. It is compatible with H.261, H.263, MPEG-2 and MPEG-4. It is a standard created without considering relevance. Thus, MVC and MPEG-4 have different characteristics.

As the biggest feature of MVC, MVC removes images obtained from multiple cameras by eliminating time redundancy as well as existing video compression standards, as well as removing inter-camera redundancy by using correlation between cameras. Since the data is compressed, all the information obtained by compressing the images obtained from several cameras is contained in one bitstream. In addition, MVC performs integer DCT in 4x4 block units, while MPEG-4 performs DCT in 8x8 block units. In addition, unlike MPEG-4, which performs motion prediction in units of 1/2 pixel in Motion Compensation (MC), MVC performs motion prediction in units of 1/4 pixels in order to increase accuracy. In addition, MPEG-4 performs spatial prediction in MVC, compared to performing AC / DC prediction in intra prediction. The MVC block type and the MPEG-4 block type used for motion estimation (ME) are also different. In entropy coding, MPEG-4 uses the Huffman table, but in MVC, Universal Variable Length Coding (UVLC) or Context based Adaptive Binary Arithmetic Coding (CABAC) is used. use.

1A is a diagram illustrating blocks used in motion estimation in the MVC standard. Referring to FIG. 1A, the MVC may divide a 16 × 16 macroblock into seven variable blocks, and each variable block may have a motion vector. That is, the macroblock may be divided into 16x16, 8x16, 16x8, and 8x8 blocks, and the macroblock divided into 8x8 blocks may be further divided into 4x8, 8x4, and 4x4.

FIG. 1B is a diagram illustrating blocks used in motion estimation in the MPEG-4 standard. Referring to FIG. 1B, unlike the MVC, the MPEG-4 standard uses only 16x16 blocks or 8x8 blocks. Therefore, in order to convert a video compressed by the MVC standard to MPEG-4 format, a block type conversion is required.

Hereinafter, when converting a video compressed according to the MVC standard to MPEG-4 format, how to convert the block type and motion information of MVC that cannot be used in MPEG-4 to reduce the time required for transcoding will be described in detail. do.

As a method of performing transcoding in the pixel region, there is a cascaded pixel-domain transcoding method. The cascaded pixel region transcoding method is a method of reconstructing both an input MVC view and a frame, and performing compression again according to the MPEG-4 standard on one view of the reconstructed image. However, in this case, since the motion prediction process for all macroblocks in all frames has to be performed again in MPEG-4 encoding, it is not suitable because of high conversion time and cost.

2 is a block diagram of a pixel region transcoding apparatus according to the present invention. According to the transcoding apparatus of the present invention, the block mode and motion information used in the MVC decoder 10 is reused in the MPEG-4 encoder 20 in order to reduce the transcoding time required to enable real-time transmission. The transcoding apparatus of FIG. 2 includes an MVC decoder 10 and an MPEG-4 encoder 20 to decode both the MVC bitstream into a pixel region and then convert the MPEG-4 bitstream.

The variable length decoder, that is, the entropy decoder 11, receives a bitstream compressed according to the MVC standard, and uses variable length decoding (VLD) or context based adaptive binary arithmetic decoding (CABAD). The decoded bitstream, the frame format of the frame, macroblock information, and motion information, which are currently reconstructed, are output. The frame format indicates a format such as an I frame or a P frame, and the macroblock information is information on which variable size block the macroblock uses for motion estimation. The motion information includes information about a reference picture and a motion vector, and is transmitted to the motion compensator 13 in the MVC decoder 10 and the motion predictor and compensator 27 in the MPEG-4 encoder 20. Macroblock information is also passed to the motion prediction and compensation unit 27 in the MPEG-4 encoder 20.

The inverse quantization and inverse DCT performing unit 12 of the MVC decoder 10 and the inverse quantization unit 24 and the inverse DCT performing unit 25 of the MPEG-4 encoder 20 dequantize and dequantize the entropy decoded bitstream. Perform a DCT. The loop filter 14 is a kind of low pass filter (LPF) that functions to smooth the boundaries between blocks. The frame converted to the pixel region is not only output to the MPEG-4 encoder but also stored in the frame storage unit 15 and input to the motion compensation unit 13 to be used for motion compensation. The motion compensator 13 performs motion compensation using the motion information received by the entropy decoder 11.

Meanwhile, the MPEG-4 encoder 20 converts the data converted into the pixel region into a DCT performer 21, a quantizer 22, and a variable length encoder, that is, an entropy encoder 23 according to a general MPEG-4 encoding method. To create an MPEG-4 bitstream. Motion compensation is also required in the MPEG-4 encoder 20. The motion prediction and compensation unit 27 in the MPEG-4 encoder 20 uses the motion information and the macroblock information received from the MVC decoder 10 to compensate for the motion. Perform In order to perform motion compensation, the motion prediction and compensation unit 27 uses a frame in which the DCT and the quantized bitstream are stored in the frame storage unit 26 through the inverse quantization unit 24 and the inverse DCT performing unit 25.

If the frame format of the bitstream input to the entropy decoder 11 of the MVC decoder 10 is a P or B frame (Predictive or Bi-predictive frame), the MPEG-4 encoder 20 consumes a lot of time. In order to omit the motion prediction process, the block type and the motion vector are reused in the macroblock. However, since the MVC standard has 1/4 pixel motion prediction and motion compensation, 7 variable block and interview prediction, and 11 macroblock coding modes, the MPEG-4 encoder 20 converts the block in order to reuse such information. Should be done.

3A to 3F are diagrams illustrating conversion of a macroblock encoding mode and a motion vector of a bitstream encoded according to the MVC standard into an encoding mode and a motion vector in MPEG-4.

Referring to FIG. 3A, if the MVC macroblock is an intra16x16 mode, an intra8x8 or intra4x4 mode, MPEG-4 converts to an intra mode.

Referring to FIG. 3B, if the MVC macroblock is in the inter16x16 mode and the motion vector is the same as in FIG. 3B, the MPEG-4 uses the inter16x16 mode and the same motion vector as the MVC.

Since there is an inter16x16 mode in MPEG-4, it can be mapped 1: 1. If the macroblock is in inter16x8 or inter8x16 mode and has a motion vector as shown in Fig. 3C, the macroblock is converted to inter8x8 mode. The motion vectors of the 16x8 or 8x16 blocks are divided into 8x8 blocks and mapped.

Referring to FIG. 3D, when the MVC macroblock is in the inter8x4, 4x8, 4x4, and direct 8x8 modes, the MPEG-4 converts to the inter8x8 mode.

The motion vectors in each block of the MVC macroblock are mapped to 2: 1 or 4: 1 in MPEG-4. That is, the motion vectors of two 8x4 blocks in the upper left are averaged to be the motion vectors of the 8x8 blocks in the upper left in MPEG-4. In the case of a 4x8 block, a motion vector of an 8x8 block in MPEG-4 is obtained by the same method as 8x4. Similarly, the motion vectors of the four 4x4 blocks in the upper right are averaged to be the motion vectors of the 8x8 blocks in the upper right in MPEG-4. The motion vectors at the bottom left are the same. The direct 8x8 mode at the bottom right performs the motion search within a small range and sets the motion vector of the 8x8 block at the bottom right.

As shown in FIG. 3E, when the macroblock is in the skip mode of the MVC, the macroblock is converted to the inter16x16 mode and the motion vector is set as a predicted motion vector (PMV). Here, the predicted motion vector refers to a vector obtained through a motion vector predictor that takes a median value of neighboring block motion vectors commonly used in MVC and MPEG-4 and calculates a motion vector.

As shown in FIG. 3F, when the macroblock is in the direct 16x16 mode of the MVC, the macroblock is converted to the inter16x16 mode and a motion search is performed within a small range to obtain a motion vector of the 16x16 block.

As described above, a method of performing block type conversion, finding a motion vector with respect to eight neighboring pixels around the motion vector obtained through block type conversion, and interpolation around the found motion vector We make 1/2 pixel and look up more motion vectors for these 1/2 unit pixels. In this way, an improvement in peak signal-to-noise ratios (PSNRs) can be obtained. Herein, when the MVC is converted in the direct 16X16 and direct 8X8 modes of the MVC, the motion vector search for the eight neighboring pixels may be omitted.

4 is a flowchart of the transcoding method of the present invention.

In summary, the transcoding method of the present invention receives block mode information and motion information of a macroblock constituting an image generated by the MVC compression method (110).

In operation 120, the MVC block mode is determined as one of an intra mode, an inter16x16 mode, and an inter8x8 mode. Which mode is determined in which case is the same as described above with reference to FIG.

In addition, the MVC motion vector is calculated as an MPEG-4 motion vector (130). That is, they are mapped to motion vectors of 16x16 or 8x8 blocks.

The motion estimation is performed on the pixels adjacent to the pixels of the motion vector coordinates calculated as described above, and 1/2 pixels are created around the pixels of the motion vector coordinates, and motion estimation is further performed on these 1/2 pixels. Readjust (140).

On the other hand, the above-described transcoding method can be written in a computer program. The codes and code segments that make up the program can be easily deduced by a computer programmer in the field. In addition, the program is stored in a computer readable media, and read and executed by a computer to implement the reference picture determination method and the motion compensation method. The information storage medium includes a magnetic recording medium, an optical recording medium, and a carrier wave medium.

1A is a diagram illustrating blocks used in motion estimation in the MVC standard.

FIG. 1B is a diagram illustrating blocks used in motion estimation in the MPEG-4 standard.

2 is a block diagram of a pixel region transcoding apparatus of the present invention.

3A to 3F are diagrams illustrating conversion of a macroblock encoding mode and a motion vector of a bitstream encoded according to the MVC standard into an encoding mode and a motion vector in MPEG-4.

4 is a process flowchart of the transcoding method of the present invention.

Claims (15)

(a) decoding the bitstream compressed by a predetermined first compression method and outputting data of a pixel region, motion vector information and macroblock information of each macroblock; And (b) receiving the data of the pixel region and performing encoding using a predetermined second compression method, performing motion compensation using the motion information and macroblock information. Coding method. The method of claim 1, wherein the step (b) receives the block mode information and the motion information of the macroblock constituting the image produced by the first compression method, converts the block mode into a 16x16 or 8x8 block mode, And the motion vector is mapped to a motion vector of a 16x16 or 8x8 block. The method of claim 1, wherein the first predetermined compression method is a compression method according to the MVC standard, and the second compression method is a compression method according to the MPEG-4 standard. (a) receiving block mode information and motion vector information of a macroblock constituting an image generated by a predetermined compression method; (b) converting the block mode to a 16x16 or 8x8 block mode; And (c) mapping the motion vector to a motion vector of a 16x16 or 8x8 block. The method according to claim 4, wherein (d) 1/2 pixels are made around the pixels of the motion vector coordinates obtained by performing a motion estimation on the pixels neighboring the pixels of the mapped motion vector coordinates. And performing a motion estimation to determine a motion vector. The compressed image transcoding method according to claim 4, wherein the predetermined compression method is a compression method according to the MVC standard. 5. The method of claim 4, wherein the step (b) determines the block mode as one of an intra 16x16 mode, an inter16x16 mode, and an inter8x8 mode. 8. The method of claim 7, wherein the step (b) converts the intra 16x16 mode to the intra 16x16 mode if the block mode is intra 16x16 mode, intra 4x4 mode, or intra 8x8 mode. The compressed image transcoding method of claim 7, wherein the step (b) converts the block mode to an inter16x16 mode if the block mode is an inter16x16 mode, a skip mode, or a direct 16x16 mode. The method of claim 7, wherein the step (b) converts the block mode to an inter8x8 mode if the block mode is an inter16x8 mode, an inter8x16 mode, or an interP8x8 mode. The compressed image transcoding method according to claim 4, wherein the step (c) maps an average of the motion vectors included in the block corresponding to the 8x8 block to which the motion vector is to be defined as the motion vector of the 8x8 block. A decoder for decoding the bitstream compressed by the first compression method and outputting data of the pixel region, motion information of each macroblock, and macroblock information; And And an encoder configured to perform motion compensation using the motion vector information and the macroblock information when the data of the pixel area is received and encoded by a second compression method. The apparatus of claim 12, wherein the encoder comprises a motion compensator, and the motion compensator receives the block mode information and the motion vector information of the macroblock constituting the image produced by the first compression method, and sets the block mode to 16x16. Or converting to 8x8 block mode and mapping the motion vector to a motion vector of a 16x16 or 8x8 block. 15. The apparatus of claim 13, wherein the motion compensator creates 1/2 pixels around pixels of the motion vector coordinates obtained by performing motion estimation on pixels neighboring the pixels of the mapped motion vector coordinates. And performing a motion estimation on the compressed image to determine a motion vector. The compressed image transcoding apparatus according to claim 12, wherein the first predetermined compression method is a compression method according to the MVC standard, and the second compression method is a compression method according to the MPEG-4 standard.

Images