Research On Image Segmentation Technology Based On Fuzzy Clustering Fused Texture Feature

Image segmentation is the key from image processing to image analysis. It is not only the objective basis of expression, but also related to the measured characteristic value. Original images are usually turned into abstract and compact forms by image segmentation and segmentation based target expression, feature extraction and parameter measurement. Image segmentation lays a solid foundation for the high level of image analysis and image understanding. Accordingly, researching and developing effective image segmentation technology is particularly important. In this paper, by introducting image space constraints into images as well as integrating texture features into the fuzzy clustering algorithm, the new and improved algorithms EnFCM_T and EnFCM_TD are put forward.The works have been done in this paper that focused on how to improve fuzzy clustering algorithm and its application in image as follow:First of all, I have done the simulation experiment on the FCM segmentation algorithm and the texture image segmentation algorithm for image. By analysing datas of the simulation experiment, I have seen the advantages and disadvantages of the two algorithms and from which respects to improve them.Then, because of the disadvantages of the traditional FCM algorithm that is not considering the pixel space information and is sensitive to noises, the usual practice is adding spatial information directly to the original objective function of the traditional FCM algorithm. But this operation of course reduces segmentation speed greatly and takes up a lot of space resources, because of that it has to scan all pixels for each iteration of the objective function. So I put forward the new algorithm, EnFCM_T,in this paper. But it introduced spatial information into the traditional FCM algorithm in the different way. I made spatial information and the original image linear add together, and regarded this additivity result as the original image of the subsequent operations; At the same time, I also fused the texture characteristics in the additivity result. Thus a new algorithm, EnFCMT, was put forward, and its simulation was done. This new algorithm improves performance and speed of segmentation.In fact, the clustering algorithm relies heavily on how to define the distance. So in order to fundamentally improve the performance of algorithm further, I have also improved the distance of the new algorithm, EnFCM_T, in the fifth chapter of this paper. Making local standard deviation and EnFCM_T distance superimposed. Then regarded the result as the final distance.This is the another new algorithm EnFCM_TD.Finally, through simulation experiment I compared the advantages and disadvantages of the three algorithms about FCM, EnFCM_T and EnFCM_TD. The result of the simulation experiment turned out that EnFCM_TD algorithm is the best of the three algorithms from the aspect of the segmentation precision, but its segmentation speed is in the intermediate position. The experiment also proved that the segmentation precision and speed of the EnFCM_T algorithm are in the intermediate position.