Automatic Fast Fuzzy Clustering Algorithm for Nucleus Segmentation of Color Pathology Images

Due to the repeated computation of the distance between pixels and cluster centres, the effective image clustering technique fuzzy c-means (FCM) takes a long time to process. In other words, FCM’s computational time greatly depends on image size or number of pixels. Therefore, researchers utilized histogram of the image for clustering for developing time-reduced clustering approach. However, finding the histogram for color image is very challenging. Another challenge of FCM is the selecting the required number of clusters before segmentation process. Therefore, this study proposes a novel automatic color image clustering approach called fully automatic histogram based fast fuzzy clustering (FAHBFFC) to overcome the both mentioned demerits. The proposed FAHBFFC automatically find the required number of clusters for the segmentation of the color images. It also utilized the histograms of the color channels of the images and hence, the proposed approach is very fast. Finally morphological reconstruction has been applied for membership filtering of the FCM to incorporate the noise-immunity. In comparison to other state-of-the-art algorithms for pathology picture segmentation, the FAHBFFC produces results that are competitive.