Research On Image Segmentation Based On Rough Set Theory

Rough set theory is a powerful tool to deal with the uncertain problems, and the advantage is that it doesn't require any additional knowledge except the data set. Image segmentation is the key step from image processing to image analysis, and plays an important role in the image project. The quality of segmentation directly determine the result of the analysis and estimation in terms of the information the image carries, it can be treated as the process to classify the pixels. Applying rough set to image segmentation, image is regarded as a knowledge system, in order to optimize the original algorithms for image segmentation.In this thesis, the basic concepts of rough set theory is described, several kinds of image segmentation methods are analyzed and compared, and several improved image segmentation algorithms by rough set are investigated. According to rough set, the inside and outside boundary factors for the image is proposed to accomplish image edge detection, roughness is adopted to select the optimal structure element for mathematical morphology operators, and the indiscernible relation is applied to complete the initial partition, so the appropriate initial cluster centers and cluster number are provided for C-means clustering algorithm. The experiment demonstrates the effectiveness of rough set's application to image segmentation.In the past, most of image segmentation methods only consider the gray or color information of the pixels, but ignore their spatial relations which are important in the process of image segmentation. Therefore, based on the histogram of image, the quantitative roughness is constructed using the concept of the upper and lower approximation, and the right threshold is selected by the adaptive peak selection strategy, so as to achieve the result image after segmentation. Furthermore, the H component in HSI color space is used to measure the distance between neighboring pixels, it reduces the amount of computation and improves the efficiency, but also get a good result. In addition, quantitative roughness is introduced into fuzzy C means clustering algorithm to refresh the fuzzy membership matrix, and the method is proved to be effective.