A Fast Motion Estimation Method

Abstract

The invention relates to video compression technology, in particular to a fast motion estimation method in the video compression technology. The invention proposes a motion estimation method that reduces the possibility of falling into a local optimum and basically does not affect the signal-to-noise ratio of an image. The gist of its technical scheme is: a kind of method for quick motion estimation, comprises the following steps: a. counts the probability size that the motion vector of the block in the image is equal to the predicted motion vector; b. when detecting the motion vector, according to the above steps a Detect the predicted motion vector in sequence from the probability obtained in the statistics from large to small; c. Use a threshold to judge the termination of the motion search. The invention reduces unnecessary searches, improves search efficiency, and is suitable for high-definition video coding and real-time coding.

Description

technical field [0001] The invention relates to video compression technology, in particular to a fast motion estimation method in the video compression technology. Background technique [0002] Motion estimation is one of the most important components in video coding system, and it has an important impact on the quality of coded image sequences. The motion estimation compensation technology can effectively remove the inter-frame redundancy in the image sequence so as to realize effective coding. The motion estimation and compensation technology based on block matching is adopted by many video coding standards, such as MPEG-1 / 2 / 4 and ITU-TH.261 / 263 / 263+ / 264, etc., because of its simple algorithm and easy implementation of software and hardware. The basic idea is to divide the coded image into macroblocks of the same size, and macroblocks can also be divided into smaller blocks. For each block, search for the closest matching block in its reference frame according to a certai...

AI Technical Summary

Problems solved by technology

[0003] Among the current search algorithms, the full search algorithm (FS) has the highest estimation accuracy. By searching every candidate point in the matching window, FS can find the global optimal motion vector, but its computational complexity is too high to be conducive to real-time Application; Traditional search algorithms such as three-step search (TSS), hexagonal search (HEXS), and diamond search (DS) are prone to fall into local optimum, unable to find the global optimum, and the search speed cannot be improved

At present, the search algorithms in the existing technology basically adopt the method of directly searching for matching blocks according to the matching criteria, instead of using the statistical probability of predicted motion vectors for priority estimation search, which wastes a lot of time and does not improve efficiency.

Moreover, most of the search algorithms in the prior art use a single-level threshold for termination discrimination. For example, the MVFAST search algorithm uses a fixed single-level threshold 512. This threshold may not be very good for some video sources with complex motion and texture. meet the requirements and have a great impact on the signal-to-noise ratio (PSNR) of the image due to the inability to terminate the discrimination early

Especially for video sequences with complex motion, although the number of search points is reduced, the signal-to-noise ratio of the image is significantly reduced

Images