An Adaptive Approach To Object Extraction and Tracking In Complex Image Sequences

Abstract : Based on visual perception principles, this article studies computational models and algorithms associated with the extraction and tracking of moving targets in complex images. It analyzes and tests the effects of changing relevant model parameters on image segmentation thresholds. What is different from certain conventional algorithms is that new methods make an integrated consideration of object background conditions, visual nonlinearities, interframe correlativity and differences into a two step integrated target extraction process including three criteria and a fast optimizing procedure. Object tracking makes use of binary template matching. Experimental results are given for visible light image sequences.