Research On A Modified FCM Refinement Segmentation Algorithm Based On Local Region

In the large amount of image data obtained,people are usually only interested in a certain part of the image or only need to study a specific area.In order to process and analyze it later,these regions need to be extracted from the whole image through image segmentation technology.In this way,the useful information in the image will be preserved,useless data will be deleted,the accuracy and the efficiency of subsequent process will be improved,and the workload of researchers can be greatly reduced.In recent years,image segmentation has become a hot topic in the field of image processing and computer vision.Researchers have successively proposed hundreds of image segmentation methods and applied them widely in real life,such as medical image analysis and auxiliary diagnosis,vehicle tracking and pedestrian detection,ground information acquisition and reconnaissance,face and fingerprint recognition,etc.It can be said that image segmentation technology is the preprocessing step of target detection,extraction or recognition and other related work.The quality of segmentation results will directly affect the reliability and effectiveness of subsequent process.Therefore,the research of image segmentation technology and algorithm is very importantImage segmentation techniques involve multiple methods.This paper mainly studies and improves segmentation algorithm based on fuzzy c-means clustering.In this paper,a FCM refinement algorithm based on local features is proposed,which can preserve edges and repress noise.The main contribution of our algorithm is to propose a weighted voting method based on image sub-blocks pre-classification.We take advantage of the pre-classification results as a new metric to measure the similarity of pixels,and combine the gray and spatial features in neighbor windows to vote and refine on these initial clustering results so as to optimize the classification relationships of pixels,and improve the segmentation effect of our algorithm on pixels located on weak edges or more seriously affected by noise.This article evaluates the segmentation results of the proposed algorithm in terms of subjective visual effects and objective quantitative indicators,and compares the results with eight other FCM-related algorithms.The experimental results on synthetic images,medical images and natural images show that the algorithm proposed in this paper is effective and the overall effect is better than other comparison algorithms.The algorithm proposed in this paper corrects the edge pixels that may be misclassified in global segmentation through a weighted voting process.At the same time,the regional homogeneity-based merge and noise correction processes make the algorithm better deal with the case of uneven grayscale in the same object,so the internal details and edge information of the objects in the image are preserved to the greatest extent,and improve the robustness of our algorithm to noisy images.It provides a new idea for the future research of image segmentation algorithm based on FCM.