Images Segmentation Study Based On The Rough Set Theory

Image segmentation is one of the most difficult tasks in image processing, and accurate image segmentation determinates the success or failure of the image processing, therefore it has received the people to take seriously. With the development of modern computer technology, a lot of efficient theories and methods of calculating bring up. An image segmentation based on a specific theory and method emerges when they are used in it. Such methods are often targeted, efficient and high quality, thus increasing people's attention. Rough set theory is one of them, it is an effectively tool dealing with imprecise, vague description of the targets of the mathematics, with the rough set of in-depth theoretical research, more and more rough set theory is applied in the image processing areas.This paper presents two new gray-scale image segmentation algorithms:One is image segmentation algorithm based on particle swarm optimization (PSO) and entropy of rough set based on boundary region; the other is image segmentation algorithm using Monte Carlo method and rough entropy standard.The first algorithm applies the rough set rough entropy based on boundary areas as the evaluation function, through the PSO searching to find the gray level corresponding the most greatly rough entropy, and using it as the best threshold to segment the image, this approach reduces the sensitivity of the algorithm on the sub-piece size, and to some extent reduces the run time of the algorithm. The second one concentrates on reducing the running time of the algorithm, which firstly uses Monte Carlo method to randomly select sample of sub-pieces to replace the entire sub-pieces of the image, greatly reduces the amount of computation, and so reduces the running time of the algorithm. It adopted a general sense of rough entropy for the evaluation function, by the way of exhaustiveness to identify the gray level corresponding the most greatly rough entropy, and get the best threshold. Two algorithms are test by MATLAB simulation, and the effectiveness and feasibility of algorithms are shown in these tests.