A Study on Performance Measure Evaluation of Semi Supervised Image Segmentation Techniques

Abstract

In image processing, segmentation means dividing the image into homogeneous regions. The task of recognizing the patterns is a crucial task. Since the result is based on which image segmentation algorithm the application is using. A number of image segmentation algorithms are available using data mining techniques like clustering based algorithms, classification based algorithms and semi supervised based algorithms. This paper discusses about two semi supervised image segmentation ideas with one standard model based algorithm (EM Cluster Algorithm). In semi supervised method both labeled and unlabeled data are used to improve the performance of segmentation. The first paper discuss about standard EM algorithms The second paper discuss about semi supervised image segmentation using mouse clicks as prior information and the third paper discuss about optimal seed selection with semi supervised segmentation. The result of analysis shows that the optimal seed selection (method III) gives better results and then clustering gives more accurate results. The Image attributes are intensity and color.