Comparative Study of the EM Color Image Segmentation in Multiple Color Spaces

摘要

The task of image segmentation in imaging science is to solve the problem of partitioning an image into smaller disjoint homogeneous regions that share similar attributes. The novel technique of the expectation-maximization (EM) algorithm with adaptively selecting dominant components in color spaces is studied as a means of improving the EMbased segmentation of color images. And simultaneously, the final segmentation is completed by a simple labeling scheme. Then the comparative study of the refined EM algorithm is done in multiple color spaces. The experimental results illustrate that the improved EM algorithm has good segmentation results with fine adaptability in RGB, CIE XYZ, and YIQ (NTSC) color spaces where the results of test image changes little. Nevertheless, these color spaces, i.e. YCbCr, HSV, CIE L*a*b*, and hlh2h3, produce poor segmentation on the reliability and accuracy of a set of test images by performance analysis with evaluation indicators.