HEVC Video Coding Framework
The HEVC coding framework is similar to the H.26X standard coding framework, and is based on a mixed block model.
Figure 1 HEVC Coding Framework
1. First, the video encoder divides the input video image into a non-overlapping encoding unit.
2. Then the predictive coding is used, the spatial correlation and temporal correlation of the video are used, and the space redundancy information is obtained by intra-frame prediction and inter-frame prediction respectively.
3. Then the predictive image block is compared with the original image block to get the predicted residual block, and then the discrete cosine transform (DCT) and quantization of the predictive residuals are obtained to obtain the quantization DCT coefficients.
4. Finally, the quantization DCT coefficients are entropy coded to get the compressed code stream.
HEVC incorporates a number of key technologies to improve performance, such as a quadtree-based coding unit, predictive-oriented intra-frame prediction techniques, motion merging techniques and advanced motion vector prediction models for inter-frame prediction, and high-precision motion compensation techniques, The de-block filter and pixel adaptive compensation technology are used to improve the quality of reconstructed images.
HEVC Coding Unit
HEVC Coding Unit has the same concept and function as the macro block in H. m, but the fast segmentation of HEVC coding is more flexible. The coding unit CU adopts four-fork tree structure, first divides a frame image into several rectangular blocks of a certain size, each block is the maximum coding unit (LCU). Each LCU can be divided into Cu of different sizes from 64*64 to 8*8, and the maximum/minimum value of CU can be modified in the configuration file.
CU adopts four-fork tree segmentation, and the specific segmentation process is marked by two variables: Split depth (Depth) and split marker (Split_flag).
HEVC Video Coding Technology