Statistic segmentation of dynamic images in video tracking systems

Abstract

One of the most difficult tasks in analyzing dynamic scenes is image segmentation on the composing regions, belonging respectively to the target and the background. The segmentation occupies the special place in the operation of the video tracking systems (VTS) due to the capability of the target detection in a search mode on the one hand, and how the quality of the segmentation affects the target location measurement precision on the other. Well known approaches to the image segmentation are the pixels intensities histogram features analysis, and the edge extracting. An advantage of such an approach is its obvious simplicity, clear physical meaning of the selection threshold and small volume of the required a priori information about segmented images.