Witness the development of H.265

Analysis of Coding Tools in HEVC Test Model (HM 1.0) – Coding Structure

2010-11-30 H.265/HEVC View Comments Views(18,782)

The HEVC test model (HM) still belongs to block-based hybrid video coding framework, except that the macroblock is extended to larger size (up to 64×64) compared with H.264/AVC. In order to facilitate syntax representation of block hierarchy, three block concepts are introduced: coding unit (CU), prediction unit (PU), and transform unit (TU). The overall coding structure is characterized by the various sizes of CU, PU and TU in a recursive manner, once the size of largest coding unit (LCU) and the hierarchical depth of CU are defined.

CU, the basic coding unit like the macroblock and sub-macroblock in H.264/AVC, can have various sizes but is restricted to be a square shape. Given the size and hierarchical depth of LCU, CU can be expressed in a recursive quadtree representation adapted to the picture. Figure 1 (b) shows maximum possible recursive CU structure in HM (LCU size = 64 and hierarchical depth = 4). In HM, the coding unit tree structure is limited from 8×8 to 64×64 for luma, that is to say, no splitting is allowed for CU3. Although the maximum possible recursive CU structure is restricted in HM, arbitrary combination of LCU (8<=LCU<=64) and hierarchical depth (<=4) can be defined to adapt to various picture resolutions. For example, if the LCU size = 16 and hierarchical depth = 2, then this is a similar coding structure to macroblock and sub-macroblock in H.264/AVC.

CU_PU

Figure 1. (a) Possible PU splittings for skip, intra and inter in HM, (b) Maximum possible recursive CU structure in HM (LCU size = 64, max hierarchical depth = 4)

Once the splitting of CU hierarchical tree is done, the leaf nodes CUs can be further split into prediction units (PUs). Like the H.264/AVC, two different terms are introduced to specify the prediction method: PU type and PU splitting. Different PU splitting corresponds to different PU types, which consist of skip, intra and inter. Figure 2 (a) shows possible PU splittings for skip, intra and inter for a leaf node CU of size 2Nx2N in HM. It is worth to note that in original proposal JCTVC-A124 four asymmetric splittings (2NxnU, 2NxnD, nLx2N and nRx2N) are also defined for inter prediction type, but this feature is not included in current HM and still under further inverstigation. All information related to prediction is signaled on a PU basis, such as intra prediction direction, motion vector difference and reference index.

In addition to the CU and PU definitions, the transform unit (TU) is defined for transform and quantization. In TU structure representation, residual quadtree structure is adopted in HM. The same maximum quadtree depth is applied for both luma and chroma components of each CU. It should be noted that the size of TU may be larger than that of the PU but not exceed that of the CU. TU have different splittings for low complexity (LC) and high efficiency (HE) configurations. For LC configuration, there are only two possible splittings from a CU to TUs depending on the value of syntax element split_transform_flag. That is to say, the residual quadtree structure is restricted to 2 levels in LC configuration. However, for HE configuration, the splitting from a CU to TUs is done recursively depending on the value of split_transform_flag, but restricted to maximum 3 levels of the residual quadtree and minimum supported transform size. The syntax element split_transform_flag may not be present in the bitstream, which should be inferred according to the coding context, such as intra coding or inter coding modes, low complexityor high efficiency cases. Please refer to JCTVC-C319 for details of the inferring process of split_transform_flag.

Table 1 summarizes the possible combinations of CU, PU and TU based on the value of split flag, PU type, PU splitting, TU for different configurations in HM.

Table 1. Possible combinations of CU, PU and TU in HM

CU_PU_TU

[References]

[1]. JCTVC-A124, Video coding technology proposal by Samsung (and BBC)

[2]. JCTVC-C311, Fast intra encoding for fixed maximum depth of transform quadtree

[3]. JCTVC-C319 , BoG report: residual quadtree structure

Permanent Link: Analysis of Coding Tools in HEVC Test Model (HM 1.0) – Coding Structure

Post Comment

*
To prove you're a person (not a spam script), type the security word shown in the picture.
Anti-spam image