Contour Tracking Based on Online Feature Selection and Dynamic Neighbor Region Fast Level Set

摘要

A novel contour tracking algorithm is proposed in this paper. The algorithm first coverts the input image into sixteen different color feature spaces and adopts Fisher discriminating rule to adaptively select the top-ranked three discriminative feature spaces who can discriminate the target region and its neighbor background region best, then adopts the nearest neighbor decision method to construct the velocity expression of fast level set only in target region and its neighbor background region but not the whole image plane, finally, contour tracking is realized by evolving the zero level set curve using dynamic neighbor region fast level set algorithm which is proposed in this paper. Experiments show that this algorithm can track target contour under conditions of moving background, illumination variation, partial occlusion and the scale and shape change of target.