Study On Image Segmentation Based On A Newlyimproved Sectional Set Fuzzy C-Means Algorithm

Image segmentation is an essential step in image interpretation and computer vision,it is also the bottleneck and hotspot in computer vision system. Today, more and more new theories are employed in image segmentation; Fuzzy C-Means Algorithm (FCM) and Genetic Algorithm (GA) are the important ones of them. Fuzzy C-Means Algorithm belongs to the unsupervised classifying method; it can avoid being inferfered by outer factors and reflect original info in image when there is no transcendental information. The general ability of electing excellent ones is the notable trait in Genetic Algorithm. In this article, studying the theories of image segmentation, Fuzzy C-Means Algorithm and Genetic Algorithm is the basis; investigation and analysis of Sectional Set Fuzzy C-Means (SSFCM) Algorithm is the core. The theory of SSFCM is more accordant with the genery principle of classifying, but the experimental results illustrate that this algorithm need to be improved. The fist betterment is based on feature space mapping and improved parameter "m ". And the experimental results illustrate that it's more efficient than traditional FCM and unimproved SSFCM. This improved method adopts Histogram to initiaize, but one Histogram can represent several images, it can not display the characteristics of every different image. So the second method employs Adaptive Genetic Algorithm (AGA) for betterment. The experimental results illustrate that AGA based SSFCM's performances on GC(Gray-leved Contrast), UM(Uniformity Measure), SM(Shape measure) and efficiency are to some extend inproved, according to the traditional FCM and the unimproved SSFCM.