5:H.264: Video Coding for new development

H.264: Video Coding for new development JVT (Joint Video Team, Joint Working Group on the video) in December 2001 the establishment of Pattaya in Thailand. By ITU-T and the International Organization for Standardization ISO two video encoding on the composition of the experts. JVT work goal is to develop a new video coding standard in order to achieve high video compression ratio, high image quality, good adaptability of the network objectives. At present, the work of the JVT has been accepted by ITU-T, the new video coding standard known as H.264 compression standard, which was also accepted by ISO, known as AVC (Advanced Video Coding) standard, is the MPEG-4 Part 10. H.264 standard can be divided into Three:

The basic level (the simple version of the application of a wide range); The main level (using a number of improve image quality and increase the compression ratio of technical measures can be used for SDTV, HDTV and DVD, etc.); Expansion level (can be used for a variety of network transmission of video streams).

Not only H.263 and H.264 than MPEG-4 of the 50 percent savings rate, but also has better network support. It into IP packets for encoding mechanism is conducive to the packet transmission network to support network streaming video transmission. H.264 has a strong anti-BER characteristics, can be adapted to the high rate of packet loss, a serious interference in the video transmission channel. H.264 support for various network resources under the classification code transmission, and thus obtain a smooth image quality. H.264 can adapt to different video transmission network, the network good affinity.

1, H.264 video compression system Compression standard H.264 video coding system layer (VCL) and network abstraction layer (Network Abstraction Layer, NAL) is composed of two parts. VCL, including VCL and VCL encoder decoder, the main functions of the video data compression encoding and decoding, which includes the motion compensation, transform coding, entropy coding compression unit. NAL is used for the VCL has nothing to do with the network to provide a unified interface, which is responsible for video data package after package to send in the network, it uses a unified data format, including a single byte of the header information, a number of words section with the group of video data frames, logical channel signaling, timing information, the end of signal sequence. Header contains the type of store signs and markers. Store signs used to indicate the current data does not belong to be a reference frame. The type of symbol used to indicate the type of image data. VCL can be transmitted by the network to adjust the current encoding parameters.

2, H.264 characteristics H.264 and H.261, H.263, the DCT transform coding is the increase in the use of DPCM coding of the difference, that is, hybrid coding structure. At the same time, H.264 hybrid coding in the framework of the introduction of a new coding method to improve the coding efficiency, closer to practical application. H.264 is not complicated options, but try to be brief the "return to basics", it is better than H.263 + + the compression performance, but also has to adapt to a wide range of channel capacity. H.264 target a wide range of applications, to meet a variety of different rate, video applications on various occasions, has good anti-error and anti-handling capacity of packet loss. H.264 basic system without the use of copyright, the nature of open, well adapted to IP and wireless networks use the Internet for the current transmission, multimedia messaging, mobile broadband network to transmit information of great significance to all. Although the basic structure of H.264 encoding with H.261, H.263 is similar, but it has made improvements in many areas, are listed below. 1. A variety of better motion estimation High-precision estimates H.263 is used in half pixel is estimated that in the further use of H.264 in the 1 / 4 pixel or 1 / 8 pixel motion estimation. That is, the real movement of the displacement vector may be based on 1 / 4 or 1 / 8 as the basic unit of pixels. Obviously, the motion vector accuracy of the higher displacement, the smaller the residual error frame, the lower the transmission rate, that is, the higher the compression ratio. H.264 is used in the 6-order FIR interpolation filter to obtain 1 / 2 pixel position value. When 1 / 2 pixel values obtained, the 1 / 4 pixel value can be obtained through linear interpolation, For 4:1:1 video format, the brightness signal 1 / 4 pixel accuracy corresponds to the color part of the 1 / 8 pixel motion vectors, it signals the need for color 1 / 8 pixel interpolation operator. In theory, if the accuracy of motion compensation doubled (for example, from whole-pixel precision to 1 / 2 pixel accuracy), can 0.5bit/Sample the coding gain, but to verify the accuracy of motion vectors found in more than 1 / 8 pixel , the system is basically there is no obvious gain, so that in H.264, only used 1 / 4 pixel accuracy motion vector mode, rather than 1 / 8 pixel accuracy. Multi-mode macroblock is estimated breakdown The forecasting model in H.264, a macroblock (MB) can be divided into seven kinds of different size, this multi-mode flexible, subtle delineation macroblock is in line with our image in the shape of the actual movement of objects, so in each macroblock may contain 1, 2, 4, 8 or 16 motion vectors. Multi-parameter frame is estimated In H.264, the frame can be a number of parameters of motion estimation, that is in the encoders of the cache code there is more than just a good parameter frame, from one encoder to choose a better coding results are given as parameters of the frame, and pointed out that the frame which was used to predict, so that you can use than just a good frame just encoded frame as a better predict the effect of the code. 4 Integer Transform '2. Small size 4 4, as the transform block size has become smaller, moving objects on the more precise delineation. This case, the image transform in the process of computation, and edges of moving objects in the convergence of error has been greatly reduced. '8. H.264 is used in the small size of the 4 'video compression coding unit used in the past 8 Transformation 2. 'DC coefficient of the four (one for each small piece of a total of four DC coefficients) for 2' 4 Transformation of chroma data 4 4 '4 of the DCT coefficient of the second 4' when the image There are large smooth areas, in order not to have a small size due to change brought about by inter-block differences in gray, H.264 Intra macroblock of luminance data 16 4 H.263 not only image transform block size has become smaller, and the transformation is an integer operation, rather than real computing, namely, encoders and decoders transform and inverse transform of the accuracy of the same, there is no "anti-conversion error." 3. More accurate intra-prediction The four are available for each pixel 17 of the nearest previously encoded pixels and the different weights for intra-prediction. 'In H.264, each 4 4. VLC unified H.264 encoding on the entropy in two ways. Unified VLC (that is, UVLC: Universal VLC). UVLC use the same code table for encoding, while the decoder can easily identify the code word prefix, UVLC bit error in the event of rapid access to re-sync. Content adaptive binary arithmetic coding (CABAC: Context Adaptive Binary Arithmetic Coding). UVLC its slightly better coding performance, but higher complexity.

3, the performance advantage H.264 and MPEG-4, H.263 + + coding performance comparison using the following six test rate: 32kbit / s, 10F / s and QCIF; 64kbit / s, 15F / s and QCIF; 128kbit / s, 15F / s and CIF; 256kbit / s, 15F / s and QCIF; 512kbit / s, 30F / s and CIF; 1024kbit / s, 30F / s and CIF. The test results indicate, H.264 than MPEG and H.263 + + more excellent PSNR performance. PSNR than H.264 in MPEG-4 high average 2dB, than H.263 + + average to high-3dB. Fourth, the new fast motion estimation algorithm New Fast Motion Estimation Algorithm UMHexagonS (China Patent) is a computational complexity compared to H.264 Zhongyuan fast full search algorithm and some savings of more than 90% of the new algorithm, called the whole "non-symmetrical cross-shaped multi-level six - shaped grid search algorithm "(Unsymmetrical-Cross Muti-Hexagon Search)", which is a whole-pixel motion estimation algorithm. because of its large sport in high-bit-rate image sequence coding at a better rate-distortion performance to maintain the conditions , the computational complexity is very low, has been formally adopted H.264 standard. ITU and ISO joint development of the H.264 (MPEG-4 Part 10) may be broadcast, communications and storage media (CD DVD) to become a unified standard, is most likely to become a broadband interactive new media standards. China's source coding standard has not yet been formulated, paying close attention to the development of H.264, the development of our source coding standards are being stepped up. Standard H264 video compression technology to enable movement to rise to a higher stage, at a relatively low bandwidth to provide high-quality H.264 video transmission is a bright spot in the application. The popularization and application of H.264 video terminals, gatekeepers, gateways, MCU, such as higher system requirements, will greatly promote the video conferencing software and hardware equipment in the continuous improvement in all aspects.