HEVC Learning II CTU, CU, CTB, CB, PB, TB

In the H264 standard, the core of the coding layer is the macro block, a macro block size of 16x16, a 16x16 brightness block, and two 8x8 chroma blocks for commonly used 4:2:0 sampling formats. corresponding to the similar structure in HEVC is the coding tree Unit (CTU), its size can be specified by the encoder, the maximum can be supported to 64x64, the minimum can be supported to 16x16. For high-resolution video encoding, using a larger size of CTU can achieve better compression performance. Here is the largest 64x64 size of CTU as an example of the HEVC coding tree structure.

HEVC first divides the entire frame image according to the size of the CTU set by the encoder, and each CTU size is 64x64, similar to H264 's 16x16 macro block division.

An Encoding Tree unit (CTU) contains a luminance-coded tree block (CTB) and two chroma-coded tree blocks (CTB) at the same location, along with some corresponding syntactic elements. , according to the size of the CTU can be 16x16,32x32,64x64, the size of the brightness CTB can also be 16x16,32x32,64x64, and always correspond to the size of the CTU, where the brightness ctb size of 64X64, Chroma is 32x32.

In HEVC coding tree block CTB can be directly as a code block CB, can also be further divided into four-fork tree form into a number of small code block CB, so in HEVC encoding block CB size is changeable, in the case of setting the size of the CTU 64x64, One brightness CB maximum is 64x64 that is a CTB directly as a CB, the minimum is 8x8, then the chroma CB largest is 32x32, the smallest is 4x4. The large CB can make the coding efficiency of the flat area greatly improved, and the small CB can make the complicated image prediction more accurate in the detail part. A luminance CB and two Chroma CB, as well as some related grammatical elements together form an encoding unit (CU).

An image based on the size of the CTU set by the encoder, can be divided into a number of non-overlapping CTU, within CTU, using a Quadtree-based loop hierarchy. Coding units at the same level have the same split depth. A CTU can have a value that contains a CU that is not divided or can be divided into multiple cu. Each CU contains a prediction unit (PU) and a transformation unit (TU) associated with it. For a 2nx2n cu mode, there are two 2nx2n and NXN for the selectable mode of the intra-Frame prediction Unit, the Inter-frame prediction mode is 8, the 4 symmetric mode 2nx2n,nxn,nx2n,2nxn, 4 Central African symmetric mode 2nxnu,2nxnd,nlx2n,nrx2n (where u,d,l represents the upper and lower left and right four directions), where nlx2n,nrx2n is divided by the ratio of left to right 1:3,3:1, 2nxnu,2nxnd is divided by the ratio of 1:3,3:1 to upper and lower. There is also a mode of skip mode, is one of inter-frame prediction, when the need to encode the motion information only the motion parameter set index (using motion synthesis technology), encoding residual information does not need to encode. A predictive unit PU contains a luminance prediction block PB and two chroma prediction blocks PB, as well as some related grammatical elements.

The transformation unit is the unit of independent transformation and quantization, and its size is also flexible. HEVC can support a 4x4-to-32x32 encoding transformation, and the base unit is a transform unit (TU). The size of the transform unit depends on the CU pattern, a CU interior, which allows TU to span multiple pu and divide it into four-tree form cabinets. The large TU can concentrate the energy better, and the smaller TU can retain more detail and a flexible way of partitioning. A transform unit tu contains a luminance transform block TB and two chroma transform block TB, and some related grammatical elements.

HEVC Learning II CTU, CU, CTB, CB, PB, TB