Motion estimation device and method of video coding system

Abstract

The invention relates to a motion estimation device in a video encoding system and a method, wherein the method comprises the following step: generating a prediction motion vector, utilizing the prediction motion vector to interpolate the reference module to get a prediction target module, calculating the absolute error sum (SAD) between the prediction target module and the target module, comparing the absolute error sum SAD with the threshold value which is set, determining whether to assess the motion of integer pixel and sub-pixel on the basis of the f the comparison result, the evaluation of the motion of integer pixel and sub-pixel comprises the following steps: firstly, searching the optimum integer pixel module through a diamond search algorithm, secondly, utilizing a sub-pixel tosearch to get the optimum sub-pixel block. The method can reduce the searching times of the integer pixel and the sub-pixel.

Description

technical field [0001] The present invention relates to a video coding system, in particular to a motion estimation device and method thereof. Background technique [0002] The two-dimensional digitized image sequence is usually expressed by the YuV three-component of pixels. Since the number of pixels in an image is very large, a large amount of data is required to represent it, which brings great challenges to the storage and transmission of images. Trouble, this requires image data to be compressed to reduce its data volume. In fact, continuous image sequences have a very high correlation in space and time, and this feature can be used to eliminate data redundancy to obtain effective data compression. [0003] The method of eliminating spatial redundancy is intra-frame prediction. In order to encode a certain area of ​​a frame image, the information of the adjacent part of this frame image and the area can be used for prediction. The key technology to eliminate temporal...

AI Technical Summary

Problems solved by technology

[0014] The algorithm proposed in the patent CN 1440203A needs to do a full search, which is not helpful to reduce the computational complexity; the patent CN 1177483C proposes a method for predicting the motion vector search, but its threshold is fixed and cannot be dynamically adjusted, so it is also It cannot effectively reduce the number of integer and sub-pixel searches

Images