background
The VCEG from ITU-T was initiated in 1998, and 2001 and MPEG proposed JVT (Joint Video Team), which was released in 2003. H.264 is also known as MPEG-4 AVC
Unlike MPEG-4 standards, H.264 focuses on improving the encoding efficiency of video signals and improving network transmission affinity H.264 application scenarios: digital TV broadcast, video real-time communication , Network video streaming media, DVD video storage, video on demand, etc.
The first version of H.264 supports 8Bit / Sample, 4: 2: 0 chrominance sampling, mainly for most universal video processing and transmission, unprocessed special applications
Special application possible demand:
Source video data accuracy exceeds 8bit / sample chroma sampling 4: 2: 2 or 4: 4: 4 ultra high yield rate and super-resolution encoding ultra-high fidelity, or part lossless encoding ...
In order to solve the needs of these professional application scenarios, 2004 JVT announced "FidenLity Range Extension", ie H.264 Frext
H.264 Video Coding Method Introduction Overall Coding Framework, H.264 still uses a mixed encoding frame of block structure, and the entire structure can be divided into network abstraction layer (NAL) and video encoding layer (VCL) each frame is divided into one. Or multiple strips Slice are encoded; each strip contains multiple macroblocks (MB, MACROBLOCK) macroblocks are H.264 basic coding units, this structure contains a 1616 brightness block + two 88 chroma blocks + other Some macroblock information
The macroblock encoding of the video coding layer VCL network abstraction layer (NAL) H.264 provides a more flexible encoding method that each macroblock is divided into a plurality of different sizes of sub-blocks for predicting the block size of the frame. * 16 or 4 * 4 Interpreter / Motion compensation can have 7 different shapes: 16 * 16, 16 * 8, 8 * 16, 8 * 8, 8 * 4, 4 * 8 and 4 * 4 A macroblock segmentation method for transforming block size for predictive residual data is 4 * 4 or 8 * 8, more sub-division, provides higher prediction accuracy and coding efficiency efficiency entropy encoding method: CAVLC and Cabac
The encoding tool used by H.264 intra prediction: H264 employs a pixel block-based intra prediction technique for reducing pixel correlation inter prediction within the image space: in the pre-standard of H.263, MPEG-4 The intra prediction data is implemented by the transform domain, and the H.264 uses the left and upper adjacent pixels of the spatial domain to predict the currently encoded pixel value transformation and quantization encoding:
If a macroblock of H.264 is an intra macroblock, its brightness component has two segmentation modes: a 16 * 16 pixel block or 16 4 * 4 pixel blocks
For each 4 * 4 pixel block, nine prediction modes are defined for 16 * 16 pixel blocks, together define a macroblock containing two 8 * 8 chroma components in one macroblock of the four prediction mode H.264, to chroma Component defines four prediction modes, the same 16 * 16 brightness components
Lossless entropy encoding other technologies
Four prediction modes of 9 prediction modes 16 * 16 pixel blocks in intra prediction of 4 * 4 pixel blocks
Inter Bloom prediction H.264 performs inter-frame prediction types of time redundancy information H.264 supported by motion compensation predictions:
Single Interpreter Prediction: P SLICE Bidirectional Inter Bloom Prediction: B SLICE
The inter prediction method of H.264 is similar to the pre-standard method such as H.263, and a certain improvement is carried out in a specific algorithm.
More block segmentation mode: higher motion vector accuracy from 16 * 16 to 4 * 4: brightness MV can reach 1/4 pixel accuracy ...
H.264's frame prediction macroblock MV is higher than the accuracy of the sub-pixel interpolation interpolation, you can achieve higher precision motion estimates
Most of the energy of integer transformation and quantization images are concentrated in the low frequency area, converting image transformation to frequency domain processing to reduce the dynamic range of image coding, effectively reducing the code rate; 4 * 4 of image / residual data in H.264 The quantization method of the integer DCT transform H.264 is quantified using scalar. In the process of quantifying / in-quantization, the quantization parameter QP determines the value of the quantization step, QP is doubled, and the quantization step is doubled. The larger the quantization step, the smaller the encoding length, the higher the compression ratio, but the greater the information loss. QP in H.264 can take 0 to 51.
Lossless entropy encoding entropy encoding utilizes the statistical characteristics of the source, no information loss during the encoding process, that is, the generated code stream can restore the original data without distortion; the classic entropy coding algorithm: Hafman code, Shannon - Fee The H.264, such as the Nuoca code, specifies different entropy encoding algorithms for different syntax elements:
UVLC (Universal Variable Length Coding): Mainly used index Columbus Coding CAVLC (Context Adaptive Binary): Context Adaptive Binary Arithmetic Coding: Context Adaptive Binary Coding Coding
Other technical loop deck filter frame / field encoding Si / SP frame code rate control ...
Our other product:
