A fuzzy region dissimilarity measure using feature space information

An inter-region color dissimilarity measure is proposed that utilizes the basic principles of region based segmentation and fuzzy clustering techniques. This method operates on the features associated to the initial image partitioning produced by watershed analysis. The subtractive clustering algorithm is employed to estimate the number of clusters and the fuzzy c-means classification method follows. The membership values assigned to each region along with a fuzzy (dis)similarity measure are used to estimate the cost between the regions. The process is completed using the shortest spanning tree merging algorithm. The proposed method is also compared to other related approaches.