Automatic Grayscale Classification using Histogram Clustering for Active Contour Models

Abstract

The problems of image segmentation using active contours are the minimization of energy criterion, involving both edge and region functional. Automatic initialization of level set function in geometric active contour model makes the process fast and convergence towards the object boundary in minimal iterations. A new technique of automatic region classification using clustering method is proposed in this paper. The approximate region count can be estimated using histogram peak finding method. The elbow statistical method is used to find the number of region using mean and standard deviation measures. The knee point in the elbow gives the accurate number of cluster. Automatic seed points of the clusters are created by Euclidian distance measure. A promising retrieval performance is achieved especially in particular examples.