A new video object motion estimation technique, based on evolutionary programming

Abstract

In order to achieve higher video coding efficiency, a new motion estimation and compensation technique has been developed, based on an evolutionary programming algorithm written in C++, that considers not only video object translations, but also rotations and color. The proposed method works as an optimization technique which uses co-evolutionary multipopulation strategy that processes the video objects of a previously segmented initial image of a video sequence, in order to build as many video object motion phenotype (VOMP) populations as main video objects are defined from the segmentation process. The designed and tested algorithm considers only one video object per planar image layer. A (VOMP) is a strings of concatenated parameters, representing predicted video object motion coefficients color information (R,G,B) and fitness values. A test frame is build with the aid of one phenotype string from each of the VOMP populations. Simulations with synthetic images have shown very encouraging results with the proposed motion estimation and compensation algorithm.