KR100733991B1 - An MPEG2-to-H.264 Transcoding Method - Google Patents

Abstract

The present invention uses motion vector information, a macroblock type, and a coding block pattern included in an MPEG2 bitstream as additional information in H.264 encoding to minimize image quality degradation and to reduce the complexity of a transcoder. A method of transcoding to .264, the method of transcoding from MPEG2 to H.264, decoding motion streams compressed by a predetermined compression method and outputting motion vector information, macroblock type information, and coded block pattern information. Making; And a selection step of adaptively selecting a macroblock mode that can be selected at the time of encoding using the motion vector information, the macroblock type information, and the coding block pattern information. Macroblock mode within B frame, 16 × 16 macroblock mode, 16 × 8 macroblock mode, 8 × 16 macroblock mode, 8 × 8 macroblock mode, intra 16 × 16 macroblock mode, intra 8 × 8 macroblock mode , One of intra 4 × 4 macroblock mode and skip (SKIP) mode is selected.

Transcoding, H.264, MPEG2, Macroblock Mode

Description

Transcoding method from MP2 to H.264 {An MPEG2-to-H.264 Transcoding Method}

1 is a selectable macroblock type in a P frame of the MPEG2 standard,

2 shows a selectable macroblock type within a B frame of the MPEG2 standard;

3 shows a selectable macroblock mode within an I frame of the H.264 standard,

4 is a selectable macroblock mode in a P or B frame of the H.264 standard,

5 shows a structure of a transcoder from MEPG2 to H.264 according to the present invention,

6 is a set of selectable H.264 macroblock modes according to MPEG2 macroblock type and coding block pattern;

7 is a comparison of the transcoding time of the transcoding method and the pre-decoding precoding method according to the present invention.

The present invention relates to a method of transcoding from MPEG2 to H.264.

More specifically, MPEG2 is used to reduce the complexity of the transcoder while minimizing image quality deterioration by using motion vector information, macroblock type, and coding block pattern included in the MPEG2 bitstream as additional information in H.264 encoding. To transcoding method to H.264.

In general, video transcoding refers to converting a characteristic of an encoded video stream into another characteristic. Characteristics of such a video stream may include a bit rate, a frame rate, a spatial resolution, an encoding standard, and the like. By classifying transcoding according to video standards of input and output video streams, transcoding can be divided into transcoding between the same video standards and transcoding between different video standards. Transcoding between the same video standards is generally the main purpose of bit rate reduction, which is SNR transcoding that reduces bit rate through conversion of quantization coefficients, frame rate transcoding that performs frame rate conversion, and frame rate. It can be divided into spatial resolution transcoding that performs spatial size conversion. Transcoding between different video standards is mainly for conversion of the video format itself, such as transmission between terminals supporting different video standards, and transcoding from MPEG2 to MPEG4 Part 2 is included in this.

Among the above-described transcodings in various areas, the present invention relates to transcoding between different standards, among which transcoding from the MEPG2 standard to the H.264 coding standard.

With the development of the network terminal field, the consumption of video contents in various terminals has increased, and the importance of such transcoding has increased. In particular, the development of a transcoding technique targeting the H.264 coding standard having the highest coding efficiency among the existing video standards can bring many advantages. However, the H.264 video standard, which has many differences from previous video image standards (MPEG1, MPEG2, MPEG4 Part2, etc.) in terms of coding element description, has as much complexity as high coding efficiency. Therefore, in the case of video format conversion to the H.264 video standard, the transcoder structure of pre-decoding and pre-decoding is shown the optimal performance, but there are many problems due to the high complexity.

Much research has been done to solve the problem of transcoder complexity to the H.264 video standard. Bilkowski et al. Proposed a fast transcoding scheme between H.263 intra frames and H.264 intra frames, which represent three intra-block prediction directions (DC, vertical, horizontal) and H in H.263. A method of transcoding between intra frames using correlation between four prediction directions at intra 16x16 at .264 and nine prediction directions at intra 4x4 is used to retrieve all intra prediction directions of H.264. Only about 4.2% complexity. However, there is a disadvantage that the image quality degradation up to 1dB depending on the bit rate. Also, Bilkowski et al. Proposed a fast transcoding technique between H.263 and H.264 predicted frames through motion vector reestimation, which significantly reduces the number of search points in motion estimation by 0.1 ~ A 0.5dB degradation in picture quality occurs, but it reduces the transcoding complexity by about 90%. Transcoding from the MPEG video standard to the H.264 video standard, which shows more differences in terms of the H.264 coding standard and the coding element technology than the H.263, is also actively progressing. "Kalva" introduced a rough transcoder from MPEG2 to H.264 with the transcoding efficiency between H.264 and MPEG2, and "Chen" et al. A transcoding technique to 264 is proposed. It receives the MPEG-2 encoded video stream as input and decodes it to the transform domain, and then performs transcoding by performing a transform between the 8x8 DCT basis and the 4x4 DCT basis and reducing the resolution (1/2) in the transform region. . This "Chen" transcoding technique has strength in terms of complexity by performing transcoding between the MPEG2 and H.264 video standards using different transformation techniques in the transform domain, but does not take into account loop filters performed in the spatial domain. There is a disadvantage.

The present invention has been proposed to solve the above problems of the prior art, by using motion vector information, a macroblock type, and a coding block pattern included in an MPEG2 bitstream as additional information in H.264 encoding. The aim is to provide a method of transcoding from MPEG2 to H.264 to minimize the complexity of the transcoder and at the same time.

In addition, the present invention can significantly reduce the complexity of the macroblock mode selection process in the transcoder by adaptively selecting the macroblock mode that can be selected in the H.264 encoding process using the macroblock type and the coding block pattern of MPEG2. Another object is to provide a transcoding method from MPEG2 to H.264.

In order to solve the above technical problem, the present invention, in the transcoding method from MPEG2 to H.264, by decoding a bit stream compressed by a predetermined compression method, motion vector information, macroblock type information, and coding block pattern Outputting information; And a selection step of adaptively selecting a macroblock mode that can be selected at the time of encoding using the motion vector information, the macroblock type information, and the coding block pattern information. Macroblock mode within B frame, 16 × 16 macroblock mode, 16 × 8 macroblock mode, 8 × 16 macroblock mode, 8 × 8 macroblock mode, intra 16 × 16 macroblock mode, intra 8 × 8 macroblock mode , One of intra 4 × 4 macroblock mode and skip (SKIP) mode is selected.

Hereinafter, a method of transcoding from MPEG2 to H.264 will be described in detail with reference to the accompanying drawings.

First, the transcoding method from MPEG2 to H.264 according to the present invention decodes a bit stream compressed by MPEG2 compression to output motion vector information, macroblock type information, and coded block pattern information, and then outputs the motion vector information and macro. The macroblock mode that can be selected at the time of encoding is adaptively selected using the block type information and the encoded block pattern information.

That is, any one of intra 16 × 16 macroblock mode, intra 8 × 8 macroblock mode, and intra 4 × 4 macroblock mode is selected as a macroblock mode within an I frame of H.264, and P or H.264 is selected. Macroblock mode within B frame, 16 × 16 macroblock mode, 16 × 8 macroblock mode, 8 × 16 macroblock mode, 8 × 8 macroblock mode, intra 16 × 16 macroblock mode, intra 8 × 8 macroblock mode Select one of the intra 4x4 macroblock mode and the skip (SKIP) mode.

In addition, when the 16x16 macroblock mode is selected as the macroblock mode, only the motion of the subpixel is estimated using only the motion vector information in MPEG2 of the integer pixel position.

The method of transcoding from MPEG2 to H.264 as described above will be described in more detail with reference to FIGS. 1 to 7.

First, all information necessary for decoding such as a motion vector is included in a bitstream encoded by MPEG2.

Among them, the macroblock type includes whether the macroblock uses prediction from a reference frame (Intra / Inter) and whether the value of the motion vector is (0,0) (MC / No-MC) or forward. And information indicating whether it is backward prediction or interpolated prediction, and has a different number of macroblock types according to the frame structure (I, P, B).

In addition, the coded block pattern information indicating whether each 8x8 block is coded / not-coded according to the value after quantization of the 8x8 block in the macroblock is also an MPEG2 bit. It is contained within a stream.

The macroblock type and coding block pattern of MPEG2 are related to the magnitude of macroblock prediction error, and the macroblock type and coding block pattern are H.264 macroblock modes (16 × 16, 16 × 8, 8). X16, etc.), which has a significant correlation with the distortion value in the Rate-Distortion Cost function.

By using the above correlation, the motion vector information, macroblock type, and coding block pattern included in the MPEG2 bitstream are used as additional information in H.264 encoding, minimizing image quality degradation and reducing the complexity of the transcoder. The transcoding process can be performed.

1 is a diagram illustrating a macroblock type in a P frame in the MPEG2 coding standard. Referring to the P frame, the P frame is divided into a motion estimation (MC) and a non-motion estimation (No-MC) according to the presence or absence of motion vector information. In case of motion estimation, it is divided into Intra and Inter depending on whether prediction from a reference frame is used or not.

In addition, since all values after the quantization in the macroblock have a value of 0, they are not encoded, and thus they are classified into coding (Coded) and un-coding (Not-Coded) according to whether or not the above coding is performed.

Finally, when using a new quantization coefficient used in the same manner as in the I frame, the macroblock type including the transformed quantization coefficient (Change QP) is selected. In particular, the macroblock type of motion unestimated inter-encoding indicates a case in which no information is included in encoding.

2 is a diagram illustrating a macroblock type in a B frame in the MPEG2 coding standard. The B frame has a macroblock type similar to that of a P frame, but more macroblock types by using bidirectional prediction.

In addition, the skip type in the B frame has a meaning different from that of the P frame. In P frame, if the motion vector is (0,0) and all blocks in the macroblock are uncoded, it is selected. However, in the case of B frame, the motion vector has the same motion vector value as the previous macroblock. If / Interpolated) is the same, it is skipped.

These macroblocks require very few bits to transmit because there is no correction to the additional DCT coefficients. The macroblock type in the B frame has 11 macroblock types in addition to the skip mode.

FIG. 3 is a diagram illustrating a macroblock mode selectable from I frames in an H.264 coding standard, wherein macroblocks in an I frame of H.264 use prediction only in a spatial domain, and the intraspace prediction as described above. The macroblock modes that use are intra 16x16 mode, intra 8x8 mode, and intra 4x4 macroblock mode. That is, a macroblock in an I frame can be selected in any of these three macroblock modes.

4 is a diagram illustrating a macroblock mode excluding an intra mode among selectable macroblock modes in a P or B frame in the H.264 coding standard, wherein the macroblock modes in P and B frames use intra-space prediction. It may also have a macroblock mode, and the skip (SKIP) mode is also selected as in MPEG2. Therefore, macroblocks in P and B frames can be selected from 16 × 16, 16 × 8, 8 × 16, 8 × 8, intra 16 × 16, intra 8 × 8, intra 4 × 4, and skip (SKIP) modes. have.

5 is a diagram illustrating a structure of a transcoder for transcoding from the MPEG2 to the H.264 standard proposed by the present invention. The transcoder to which the present invention is applied includes a macroblock type used in a decoding process of MPEG2, and By using the coding block pattern, the macroblock mode in H.264 is adaptively selected in the case of visual prediction along with the selection of intra and intra prediction in the H.264 encoding process.

In addition, when the 16x16 mode is selected after the adaptive mode selection, the motion vector information in MPEG2 of the integer pixel position is used, and in this case, only sub-pixel motion estimation is performed.

6 is a diagram illustrating a set of selectable H.264 macroblock modes according to an MPEG2 macroblock type and an encoding block pattern. The set of selectable macroblock modes is classified into five types according to the MPEG2 macroblock type. The MC coded macroblock type is classified into three types again.

FIG. 7 is a graph illustrating a comparison of transcoding times in the case of using the transcoder to which the present invention is applied, compared to the case of full encoding in H.264 after full decoding of MPEG2. As can be seen, it can be seen that the transcoder of the present invention can shorten transcoding time of up to 61.9% and an average of 55.4% on six standard images.

As described above, the macroblock mode selectable in H.264 encoding is adaptively selected according to the macroblock type of MPEG2, thereby transcoding from MPEG2 to H.264 having low complexity.

The present invention is not limited to the above embodiments, and various modifications and changes can be made by those skilled in the art, which are included in the spirit and scope of the present invention included in the appended claims.

As described above, the present invention can provide a motion re-estimation in the spatial domain, that is, a transcoding method in the spatial domain, and the structure of the transcoder is divided into an MPEG2 decoder and an H.264 encoder, thereby decoding MPEG2. The unit performs predecoding of the input MPEG2 compressed bitstream and the H.264 encoder adaptively selects the macroblock mode by using information of the macroblock type and the encoded block pattern as well as the motion vector included in the MEPG2 compressed bitstream. This has the effect of enabling a lower complexity transcoder.

In addition, the present invention has an image quality degradation of only 0.03 dB on average compared to pre-decoding of the MPEG2 bitstream after H.264, while reducing the overall transcoding time by up to 61.9%.

In addition, the present invention not only provides a low complexity transcoding technique having minimal image quality degradation by using additional information included in the MPEG2 bitstream, but also can be used for real-time transcoding from MPEG2 to H.264 when used with a fast motion estimation technique. There is an effect to reach more.

Claims (9)

In the transcoding method from MPEG2 to H.264, Decoding the bit stream compressed by a predetermined compression method and outputting motion vector information, macroblock type information, and coded block pattern information; And A selection step of adaptively selecting a macroblock mode that can be selected at the time of encoding using the motion vector information, macroblock type information, and coding block pattern information; The selection step, Macroblock mode in P or B frames of H.264 with 16 × 16 macroblock mode, 16 × 8 macroblock mode, 8 × 16 macroblock mode, 8 × 8 macroblock mode, intra 16 × 16 macroblock mode, intra A method of transcoding from MPEG2 to H.264, wherein any one of 8 × 8 macroblock mode, intra 4 × 4 macroblock mode, and skip (SKIP) mode is selected. The method of claim 1, The selection step, In MPEG2, any one of intra 16x6 macroblock mode, intra 8x8 macroblock mode, and intra 4x4 macroblock mode is selected as a macroblock mode in an I frame of H.264. Transcoding method to 264. delete The method of claim 2, The selection step, A skip macroblock in a P or B frame of MPEG-2 is selected as a skip macroblock mode of H.264. The method of claim 2, The selection step, The motion non-estimated intra macroblock in I, P or B frames of MPEG-2 is selected in consideration of H.264 intra 16 × 16 macroblock and intra 4 × 4 macroblock mode. Transcoding method to 264. The method of claim 2, The selection step, The method of transcoding from MPEG2 to H.264, characterized in that the motion unestimated inter macroblock in a P-frame of MPEG-2 is selected in consideration of H.264 skipping and 16 × 16 macroblock mode. The method of claim 2, The selection step, The method of transcoding from MPEG2 to H.264, wherein the motion estimation unsigned macroblock in the MPEG-2 P or B frame is selected considering only the skip of H.264 and the 16 × 16 macroblock mode. The method of claim 2, The selection step, The method of transcoding from MPEG2 to H.264, wherein the motion estimation encoded macroblock in the P or B frame of MPEG-2 is selected in consideration of the set of H.264 selectable macroblock modes according to the coding block pattern. The method of claim 2, The selection step, And 16x16 macroblock mode is selected as the macroblock mode. The method of transcoding from MPEG2 to H.264 using only motion vector information in MPEG2 of an integer pixel position is estimated.

Images