Contrast Enhancement through Clustered Histogram Equalization

This study proposed a contrast enhancement algorithm. Some methods enhance images depending on only the global or the local information, therefore it would cause over-enhancement usually and make the image look unnatural. The proposed method enhances image based on the global and local information. For the global part, we proposed mapping curves to find the new average, maximum and minimum intensity to try to suit the concept of Human Visual System (HVS) for obtaining the better perceptual results. For the local part, we utilized fuzzy c-means clustering algorithm to group image and we can obtain the information of intensity distribution and pixel number from each group. Then we calculate weights according to the information and enhance images by Histogram Equalization (HE) depending on the weights. The experiment results show that our method can enhance the contrast of image steadily and it causes over-enhancement with lower probability than other methods. The whole image not only looks natural but also shows detail texture more clearly after applying our method.