H.265/HEVC standard is a revolutionary breakthrough in video compression technology. A key requirement for the new standard is to double the data compression ratio for the same video quality compared to the wide spread H.264/AVC standard. To achieve this goal, a number of algorithmic tools are proposed to take into account many aspects of video compression.
Three abstractions are proposed for efficient and flexible representation of videos with various resolutions including Coding Unit (CU), Prediction Unit (PU), and Transform Unit (TU). A CU is a basic compression unit similar the macroblock concept used in AVC and MPEG-2 standards, but it has more flexible nature. For efficient compression of HD and 4K video the supported size of the CUs are 16 × 16 or greater, for example, 32 × 32 and 64 × 64. A CU is a prediction block and may contain several PUs Asymmetric Motion Partitioning (AMP) is used to increase the coding efficiency of a video frame’s irregular patterns. A TU is a transform unit and may include one or more PUs. It supports the standard 4 × 4 and 8 × 8 transformations and additional rapid transformations of 16 × 16 and 32 × 32.
Several new algorithmic tools are proposed to improve the coding efficiency of the inter frame prediction. Prediction of motion vectors is enhanced, and 1/4-pel interpolation filters are modified.
Additional prediction angles are included to acquire more accurate prediction structures to increase the coding efficiency of intra frame prediction by reducing the energy of the residual signal.
A combination of multiple filters is used to reduce the distortion between the original and the reconstructed video frames. In addition to a modified deblocking filter, adaptive SAO (Sample Adaptive Offset) filter is added to reduce the secondary distortions of local areas.
The proposed entropy encoding is Syntax-based context-adaptive Binary Arithmetic Coding (SBAC). SBAC is an adaptive binary arithmetic encoding method, which employs context models and provides high coding efficiency of various syntax elements with different statistical properties. The adaptive coefficients bypass (ACS) is used to increase the efficiency of entropy encoding’s transform coefficients by explicitly defining the scan order of each transform block.
Introduction of a tile, a new concept, is yet another innovation of H.265/HEVC standard compared to H.264/AVC standard. Tiles divide the frame into a grid of rectangular regions that can independently be decoded/encoded. Unlike slices, the tiles increase the compression ratio not being separate units of syntactic encoding. On the other hand, as in the case of slices in the AVC standard, the loss in compression efficiency from larger inter frame prediction error and entropy coder context update at the boundaries is negligibly small. The use of tiles opens up new possibilities to increase encoding and decoding of video data using parallel processing on modern multi-core desktop and mobile platforms.
The proposed algorithmic tools described achieve on average a bit rate reduction of approximately 40% as compared to the H.264 AVC for 720p, 50% for 1080 and up to 50% for 4K resolutions, respectively. Also, the compression ratio tends to increase for higher video resolutions, the result consistent with the proposed specifications of H.265/HEVC standard.
April 20, 2017