Automatic fuzzy contrast enhancement using gaussian mixture models clustering

Abstract

In this work, an algorithm for Contrast Enhancement is proposed using Gaussian Mixture Models to identify groups of pixels in an image corresponding to different objects, and an automatic function for generating a Fuzzy Logic Inference System. The system relates the input values of the original image to output values for a new image so that each group is processed according to their properties. The input and output fuzzy sets are created using a proposed function based on the clustering results. The optimal number of groups for the segmentation of the image is found using statistical metrics. The results were compared with the original images and the images processed using histogram equalization, considering the standard deviation as a metric for contrast. The output images presented an increment in the contrast without increasing the noise.