Parallel implementation method, device and system for advanced motion vector prediction AMVP

Abstract

The embodiment of the invention provides a parallel implementation method, a device and a system for advanced motion vector prediction AMVP. The parallel implementation method includes that: a parallel area corresponding to parallel control parameters is obtained according to the parallel control parameters set by a coding terminal user; motion vector prediction selection sets of all the interframe units are constructed in a parallel way by the interframe prediction units of a coding unit in the parallel area corresponding to the parallel control parameters; motion vector prediction is carried out in a parallel way by all the interframe prediction units in the coding unit with utilizing the motion vector prediction selection sets constructed by the interframe prediction units, and predicted motion vectors are generated separately; motion estimation is carried out by each interframe prediction unit according to the predicted motion vector of the interframe prediction unit separately. In the parallel implementation method, the device and the system for advanced motion vector prediction AMVP, motion vector prediction selection sets of all the interframe units are constructed in a parallel way by the interframe prediction units in the parallel area corresponding to the parallel control parameters, thus parallel of motion estimation for prediction units of the coder end is achieved.

Description

technical field [0001] The present invention relates to the technical field of multimedia video encoding and decoding, in particular to a parallel implementation method, device and system of Advanced Motion Vector Prediction (AMVP). Background technique [0002] In the high-efficiency video coding standard, a coding unit (Coding Unit, CU) is a basic unit of video coding. The size of the coding unit can be from 64x64 to 8x8. Among them, 64x64 becomes the largest coding unit (Largest Coding Unit, LCU). In each coding unit, do prediction, transform quantization, entropy coding, post-processing to encode. When the coding unit performs prediction, the coding unit can be divided into prediction units (Prediction Unit, PU) of different sizes, and the prediction unit is a basic unit of prediction. Coding units larger than 8x8 can support 2Nx2N, 2NxN, Nx2N three symmetrical prediction unit division methods and 2Nx0.5N, 2Nx1.5N, 0.5Nx2N, 1.5Nx2N four asymmetric prediction unit divi...

AI Technical Summary

Problems solved by technology

[0005] From the above-mentioned prediction unit encoding process, we can see a problem: the current prediction unit will use the motion information of the surrounding adjacent prediction units when encoding, and these motion information cannot be obtained before the adjacent prediction units are encoded. , so that the prediction unit must be serially completed in sequence during encoding

This makes it impossible for multiple prediction units within the same coding unit to construct their motion vector reference sets in parallel.

Images