Key words:
AVC, encoding
Code decoding technology has two sets of standards, ITS (ITU-T) standard H.261, H.263, H.263 +, etc .; also ISO MPEG standard MPEG1, MPEG2, MPEG4, etc. H.264 / AVC is the latest standard for the advantages of the two major organizations, the advantages of H.263 + and MPEG4, and the most valuable part is undoubtedly a higher data compression ratio. Under the same image quality conditions, the data compression ratio of H.264 is 2 times higher than H.263, 1.5 times higher than MPEG-4.
Here we briefly introduce the concept and development of H.264, and discuss the possibility of H.264 technology practicalization.
H.264 / AVC is?
The H.264 / AVC standard is developed by ITU-T and ISO / IEC, which is positioned to cover the entire video application, including: low-rate wireless applications, standard definition, and high-definition TV broadcast applications, on the Internet Video stream application, transmit high-definition DVD video, and high quality video applications applied to digital cameras, etc.
ITU-T is named H.264 (previously called H.26L), and ISO / IEC calls an MPEG-4 advanced video coding (AVC), and it will become the MPEG-4 standard 10 parts. Since AVC is currently expanded by the MPEG-4 standard, it will inevitably benefit from MPEG-4 development well-developed infrastructure (such as system layers and audio, etc.). Obviously, the MPEG-4 AVC of the MPEG-4 Advanced Simple Framework (ASP) will be superior to the current MPEG-4 video compression standard, which will primarily apply high compression ratios and hierarchical quality needs. direction.
ITU-T and ISO / IEC are responsible for all the customs of all international video compression standards. So far, the most successful video standard is MPEG-2, which has been widely accepted by various market sectors such as DVD, digital TV broadcast (overwriting cables and communication satellites) and digital sets. Since the MPEG-2 technology, the new H.264 / MPEG-4 AVC standard has huge improvement in coding efficiency and quality. Over time, in many existing applications, H.264 / MPEG-4 AVC will replace MPEG-2 and MPEG-4, including some emerging markets (such as ADSL video).
The evolution of digital video codec technology
International standards are usually developed by the International Standardization Organization ISO on the basis of Technical Suggestions for Telecommunication Alliance ITU. The digital video codec standard has also experienced multiple changes, and the H264 standard has risen moving image compression technology to a higher stage, providing high quality image transfer on lower bandwidth is the high-quality application highlight of H.264. The promotion and application of H.264 is higher in the system of video terminals, gatekeepers, gateways, and MCUs, which will strongly promote the continuous improvement of video conferencing soft equipment in various aspects.
H.264 core competitiveness
The most valuable part of H.264 is undoubtedly a higher data compression ratio. The basic principle of compression technology is to filter non-important information in the video file to allow data to be transmitted in the network faster. Under the same image quality conditions, the data compression ratio of H.264 can be 2-3 times higher than the MPEG-2 high of the MPEG-2 high than the MPEG-4 high than the MPEG-4 high. Because of this, after H.264 compressed video data, less bandwidth required during network transmission, and more economical.
When the MPEG-4 requires a 6Mbps transmission rate match, H.264 only requires only 3 Mbps-4Mbps transmission rate. We use transportation to do more metaphors: also use a truck to transport a large box, if MPEG-4 can lose half of the box, then H.264 can lose the box to the original 1/4, When the weight of the truck is not changed, the H.264 is twice as the MPEG-2 is doubled.
H.264 Get superior performance is a substantial increase in computational complexity, such as hierarchical design, multi-frame participation, multi-mode motion estimation, improved intra prediction, etc., significantly increased prediction accuracy, thus obtaining more than other The standard is much more compressed.
Constantly improved hardware processing capabilities and continuous optimized software algorithms are the foundation of H.264. As early as ten years ago, the CPU for several tens of trillion reached top, and now the ordinary desktop, the clue of CPU has been up to several thousand megabytes. According to Moore's law, the capacity of the chip unit is doubled every 18 months, so the computational complexity of H.264 is slightly insignificant relative to performance improvement. What's more, new computing methods are endless, and the H.264's hunger demand for processing speeds is relatively alleviated.
Comparison of H.264 and MPEG-4
In the case of a very low rate (32-128 kbps), H.264 has a performance multiplication effect compared to MPEG-4, ie the H.26L media stream of the same yard ratio compared to the MPEG-4 media stream. 26L has a gain of approximately 3 pieces (the quality level is multiplied). 32kbps H.26L media stream, its hiker is similar to 128K MPEG-4 media flow. That is, in the same picture quality, the syrup rate of H.264 is only a quarter of MPEG-4.
H.264 / AVC core technology overview
This new standard is composed of several processing steps:
Inter-frame and intra prediction
Transform (and reverse transform)
Quantization (and inverse quantization)
Loop filtering
Entropy encoding
The single picture stream forms a video that can be divided into "macroblock" of 16x16 pixels, which simplifies the processing of each step in the video compression algorithm. For example, a picture taken from the standard definition standard video stream solution (720x480) is divided into 1350 (45x30) macroblock, and then further processing in the macroblock level.
Inter-frame prediction
Improved motion estimation. Motion estimation is used to determine and eliminate time redundancy between different pictures existing in the video stream. When the motion estimation search is based on the picture in the past direction, the encoded picture is called "P frame picture", when the search is based on the two directions in the past and future, then the encoded picture is called "B frame picture" .
In order to improve the coding efficiency, the macroblock is split into smaller blocks in order to include and separated the moving macro block in the "H.264 Motion Estimation-Improved Motion Estimation). Then, the motion vector of the previous or future picture is used to predict a given block. H.264 / MPEG-4 AVC invented a smaller block, which has better flexibility, and can have higher prediction accuracy in the motion vector.
Intra prediction
If you can use motion estimates, you can use intra estimation to eliminate spatial redundancy. Internal estimates predict the current block by pumping adjacent pixels in different directions in different directions in a predefined set. Then the prediction blocks and the real blocks are encoded. This method is especially useful for H.264 / MPEG-4 AVC, especially for a flat background that often has space redundancy.
Transform
The results of motion estimates and internal estimates are converted from the spatial domain to the frequency domain by transform. H.264 / MPEG-4 AVC uses an integer DCT4x4 transform. MPEG-2 and MPEG-4 use floating point DCT8X8 transform.
The smaller block H.264 / MPEG-4 AVC reduces block effects and obvious artificial traces. The integer coefficient eliminates the accuracy loss caused by the floating point coefficient operation in MPEG-2 and MPEG-4.
Quantify
The coefficients after the transformation are quantified, reducing the predictive amount of integer coefficients and eliminating it is not easy to perceive high frequency coefficients. This step is also used to control the bit rate of the output remaining in a substantially constant constant.
Loop filtering
The H.264 / MPEG-4 AVC standard defines a decimal block filtering process for the 16x16 macroblock and 4x4 block boundaries. In this case, the purpose of filtration is to eliminate artificial traces caused by different motion estimates (such as motion estimates and internal estimates) due to adjacent macroblocks. In this case in this case, the purpose of filtration is to eliminate artificial traces that may cause differences due to transformation / quantization and from phase neighboring block motion vectors. The loop filter is modified by a content adaptive nonlinear algorithm two pixels in the same side of the macroblock / block boundary.
Entropy encoding
The quantization coefficient of 4x4 must be reordered before entropy encoding. According to these coefficients, the prediction algorithms used for motion estimates or the different scan types are selected to create a different scan type to create a reordering serialization. The scan type is sorted in the order from low frequency to high frequency. Since most of the high frequency coefficient tends to zero, then the number of zero can be reduced by using the run coding, thereby efficiently reaching the purpose of entropy coding.
Serialization of entropy coding - coefficient
The motion vector, quantization coefficient, and macroblock are indicated by the byte stream of the mapping symbol. Entropy encoding indicates frequently used symbols by designing a few bit bits, and more bitits are used to represent symbols that are not currently used.
Our other product:
