Video affine motion estimation method of adaptive factors

Abstract

The invention discloses a video affine motion estimation method of adaptive factors. The method comprises the following steps: judging a zoom factor of a current to-be-predicted macro block by using alinear weighting prediction error corresponding to a translational motion vector and 2D weighting autocorrelation of a reference frame; and secondly, keeping the zoom factor unchanged, expressing anaffine motion compensation error as a quadratic function about a rotation angle, calculating an adaptive rotation angle under the affine motion through the Vieta theorem, and then obtaining an affinemotion vector of the current to-be-predicted macro block. By adoption of the video affine motion estimation method disclosed by the invention, the "violent" parameter search or iterative solution of the traditional affine motion estimation method is avoided, the optimal zoom factor and rotation angle can be directly calculated, and the number of sub-pixel interpolation operations is significantlyreduced, thereby improving the compensation quality of the traditional block matching motion estimation method while ensuring the instantaneity.

Description

technical field [0001] The invention relates to the field of video coding and compression, in particular to a video affine motion estimation method with fast operation speed, high motion compensation quality and the ability to effectively predict the adaptive factor of the affine motion existing in the video. Background technique [0002] Motion estimation is an effective temporal prediction technique, and most of the improvements in video coding efficiency over the years have been due to better motion estimation algorithms. However, the computing resources consumed by the motion estimation link often account for more than 50% of the total resources of the encoder, and even up to 80%. In order to achieve a better compromise between complexity and prediction accuracy, existing video coding standards generally use block matching motion estimation algorithms based on translation models, and have successively proposed a variety of fast block matching strategies, such as downsamp...

AI Technical Summary

Problems solved by technology

However, the solution process of the above three methods all involves a large number of interpolation operations with sub-pixel precision, and its computational complexity is even much higher than that of the full search of the block translation model, which cannot realize real-time motion estimation / compensation, which largely limits its practicality

Images