Image Segmentation And Restoration Based On KFCM Clustering And GMP Pursuit Algorithm

With the rapid development of human science and computer technology, people ondigital image processing technology has no longer just stayed in simple image processing, inorder to meet certain needs and achieve a certain purpose, specialized image processinggradually becomes a research topic which people are eager to. Image segmentationtechnology which isolates target and background and image restoration technology whichrepairs and restores the damaged regions are two typical examples. Both technologies can notonly promote the development of in-depth study of digital image processing, but also toexpand its wide range of applications in different areas, it is important to the development ofhuman technology and the improvement of living standards.This paper firstly introduces the FCM algorithm based on kernel function (KFCM), thisalgorithm has good clustering effect for the ambiguity and uncertainty of the data set, butthere are two drawbacks: firstly, we do not know the number of categories in advance, itneeds to artificially determine. Secondly, it is more sensitive to noise. An improved algorithmbased on KFCM above two defects is presented, and applied to image segmentations. Thealgorithm firstly uses k-means algorithm to estimate the number of classes, and then useKFCM clustering algorithm to cluster. In order to reduce noise interference and improve thealgorithm noise immunity, the original membership functions are modified to make themembership of the new algorithm be the average of the neighborhood. The experimentalresults show: the new algorithm does not require artificially determined number of categories,and the ratio of the classical FCM algorithm and KFCM algorithm can better noisesuppression.Secondly, learning the basic theory of compressed sensing, this paper proposes a tracking (GMP) image restoration based on genetic matching algorithm using GMP to find the bestatoms. Using single selection operator can lead to premature and reduce the diversity of thepopulation. In order to solve this problem, firstly the elitist strategy is selected, and in order toprotect the diversity of the population, the tournament selection method is used to reduce theselection pressure, and finally uses Roulette Wheel to accelerate the convergence in order toimprove the selection pressure. The mixed improved selection operator is applied to the imagerestoration algorithm proposed in this paper, the experimental results show that: the algorithmhas certain operability and repairing effects are closer to the original image.Finally, a summary of the full text is given and further research prospects are proposed.