Developing and Studying the Algorithm for Segmentation of Simple Images Using Detectors Based on Doubly Stochastic Random Fields

Abstract

A problem of segmentation of simulated images with the simplest objects is considered. In addition, an algorithm is developed for segmentation of doubly stochastic images that is based on correlation properties rather than brightness properties of images. The efficiency of this algorithm is studied. To increase the segmentation accuracy, an anomaly-detection algorithm based on doubly stochastic random fields is proposed. The proposed algorithms are studied for various levels of signals. They are compared with the known segmentation algorithms. Image-segmentation software is developed. Its brief description is given.