CN103414891B - A kind of bottom-up fast quad-tree trimming algorithm of HEVC intraframe coding - Google Patents

Abstract

The invention discloses the bottom-up fast quad-tree trimming algorithm of a kind of HEVC intraframe coding.Have employed the coding unit dividing mode of quad-tree structure in HEVC, the present invention is according to the feature of the quad-tree structure decision making algorithm of bottom-up cutting, and design achieves a kind of fast algorithm.Select to skip the mode decision process of current layer coding unit by the coding information self-adapting of known quaternary tree bottom coding unit, thus avoid unnecessary calculating, reduce the complexity of intraframe coding.The present invention can efficiently reduce the scramble time in the case of not reducing code efficiency.

Description

A kind of bottom-up fast quad-tree trimming algorithm of HEVC intraframe coding

Technical field

The present invention relates to technical field of video coding, particularly to the intraframe coding of a kind of efficient video coding standard HEVC Fast algorithm.

Background technology

Efficient video coding standard (High Efficiency Video Coding, HEVC) is remote by International Telecommunication Union Journey communication standardization tissue (ITU-T) Video Coding Experts group (VCEG) and International Standards Organization/International Electrotechnical Commission (ISO/ IEC) video encoding standard of new generation that motion picture expert group version (MPEG) is formulated jointly is moved.VCEG with MPEG became in January, 2010 Found joint video coding work group (Joint Collaborative Team on Video Coding, JCTVC), and passed through Repeatedly JCTVC meeting discussion, study and improve, final ITU-T and ISO/IEC has formally issued HEVC mark in January, 2013 Standard, i.e. ITU-T H.265 with MPEG-H Part2.On the premise of same application condition and video quality, code check ratio is previous H.264/MPEG-4AVC video encoding standard reduces half.

HEVC have employed with previous video encoding standard H.264/MPEG-4AVC as the basic subrack of hybrid coding Frame.In conducting frame during coding, frame of video be divided into non-overlapping copies code tree block (Coding Tree Block, CTB), each CTB can be become multiple encoding block (Coding Block, CB) by quadtree decomposition.The size of CB can be 64 × 64,32 × 32 or 16 × 16.In conducting frame during prediction, CB can be one or resolve into the prediction of four equal cun of chis Block (Prediction Block, PB)；When carrying out transform and quantization, a CB can be become multiple change by quadtree decomposition again Change block (Transform Block, TB).For intraframe predictive coding, HEVC defines altogether 35 kinds of intra prediction modes, bag Include Intra_Planar, Intra_DC and 33 kinds of directional prediction patterns Intra_Angular [i], i=2 ..., 34.This spirit Partitioned organization and the more intra prediction mode lived make HEVC ratio in code efficiency H.264/MPEG-4AVC have very Big raising.But, it also substantially increases the computation complexity of coding simultaneously.

When carrying out the decision-making of partitioned organization, reference software HM of HEVC have employed a kind of bottom-up quaternary tree cutting Algorithm.First, a CTB is formed a full quaternary tree, as shown in Figure 1.The most each node is a CB, and we are from upper Under to, from left to right the position of each node of labelling be (i, j), i ∈ [0,3], j ∈ [0,4ⁱ-1].If X_i,jFor node (i, j) CB when not decomposing, C_i,jFor node, (J () is optimum during one CB of coding for i, optimum CB quad-tree structure j) Rate distortion costs.Bottom-up quaternary tree trimming algorithm is by each joint in the full quaternary tree of the order traversal of depth-first Point, whenTime, cutting node (i, subtree j), it may be assumed that the CB on this node does not carry out quaternary tree and divides Solve.

Summary of the invention

For the problems referred to above, the bottom-up fast quad-tree cutting that the invention provides a kind of HEVC intraframe coding is calculated Method.Encoder complexity can be efficiently reduced in the situation not reducing code efficiency.Although bottom-up quaternary tree cutting is calculated The method partitioned organization than exhaustive every kind of CB is more efficient, but still has higher complexity.For a quaternary tree layer The CTB of level up to 4, required for carry out frame mode search the quantity of CB beTherefore, the present invention devises one Novel fast quad-tree trimming algorithm, reaches to reduce the mesh of computation complexity by skipping some unnecessary nodes 's.

For achieving the above object, the technical solution adopted in the present invention is:

A node does not decomposes to make Q ()=0 represent, and Q ()=1 represents decomposition downwards.If Q is (C_i+1,4j+k)=1, then it represents that joint (i, containing complex texture structure in the CB of kth child node j), needs to resolve into less block and carries out infra-frame prediction point Coding, such that it is able to infer that present node is likely to needs and decomposes.Therefore, the present invention uses Decide whether to skip present node (i, frame mode search procedure j) as threshold value.Work as N_i,jDuring >=i+1, then skip current Node (i, j).As shown in Figure 2, node (1,1) is subtree, then skip the pattern search process at node (0,0) place.

Beneficial effect

The present invention realizes in HEVC reference software HM8.2, and uses intra-frame encoding mode to 5 classifications of HEVC totally 20 Cycle tests is tested.Encoder is set to acquiescence, and the quantization parameter of selection has 22,27,32 and 37.With HM8.2 Central Plains The mode decision algorithm begun is compared, and percentage ratio that the present invention saved on the scramble time and distortion performance are as shown in Table 1. It will be seen that the present invention can reach more than the saving of the scramble time of 20%, in code efficiency, only have average 0.46% simultaneously Code check increase.

Accompanying drawing explanation

Fig. 1 is full quaternary tree；

Fig. 2 is Fast trim example schematic；

Fig. 3 is quick bottom-up trimming algorithm；

Detailed description of the invention

Embodiment 1

As it is shown on figure 3, the quick bottom-up trimming algorithm that the present invention is carried uses the mode of a kind of recurrence to carry out.Wherein D represents most number of levels that quaternary tree is allowed.For the node of the bottom, directly carry out frame mode search and obtain optimum Rate distortion costs function, be designated as J (C_i,j).For the node of other level, (i, j), i < D-1, the structure of child node and rate are lost True cost is known.If $\underset{k=0}{\overset{3}{\mathrm{\&Sigma;}}}Q\left(C_{i+1,4j+k}\right)\&GreaterEqual;i+1,$ Then $C_{i,j}={\&cup;}_{k=0}^3{C}_{i+\mathrm{1,4}j+k};$ IfThen Carry out frame mode search and obtain the rate distortion costs function J (X of optimum_i,j), now, if Then C_i,j=X_i,j, otherwise $C_{i,j}={\&cup;}_{k=0}^3{C}_{i+\mathrm{1,4}j+k}.$

The increase of the percentage when average bit rate that table 1 scramble time is saved

It is last that it is noted that obviously above-described embodiment is only for clearly demonstrating example of the present invention, and also The non-restriction to embodiment.For those of ordinary skill in the field, can also do on the basis of the above description Go out change or the variation of other multi-form.Here without also cannot all of embodiment be given exhaustive.And thus drawn What Shen went out obviously changes or changes among still in protection scope of the present invention.

Claims (1)

1. the bottom-up fast quad-tree trimming algorithm of a HEVC intraframe coding, it is characterised in that:

A node does not decomposes to make Q ()=0 represent, and Q ()=1 represents decomposition downwards, as Q (C_i+1,4j+k)=1, then it represents that (i, containing complex texture structure in the CB of kth child node j), needs to resolve into less block and carries out in frame node Predictive coding, infers that present node is also required to decompose；

Use(i, frame mode j) was searched for decide whether to skip present node as threshold value Journey, works as N_i,jDuring >=i+1, skip present node (i, j).