A novel fuzzy metric for objective evaluation of interactive image segmentation algorithms based on correlation

Abstract

Image segmentation is a relevant research area in computer vision and plays a major role in a broad range of applications; therefore it is necessary to be able to evaluate the performance of image segmentation algorithms objectively. For natural images, usually we use interactive image segmentation to provide a means of accurately extracting semantic objects from an image efficiently. In interactive segmentation the users only need to roughly indicate the location and region of the object and background by using markers. In this paper we present a new fuzzy error measure index to obtain the interactive image segmentation error, based on correlation of pixels. Also this measure is based on segment-by-segment comparisons of a segmented image and a ground-truth and is sensitive to the position of each object. Experimental results were obtained for a selection of test images and showed that it's a proper measure for comparing interactive image segmentation algorithms.