Image Segmentation Based on Cumulative Residual Entropy

Cumulative residual entropy (CRE) is an essential concept in information theory and have more general mathematical properties in contrast to entropy. However, it is observed that research on CRE has relatively little consideration in image processing. Image thresholding technique plays a crucial role in several of the tasks needed for pattern recognition and computer vision. In this paper, we study, implement, and apply the CRE measure for image thresholding. Firstly, we have defined a thresholding criterion, which is based on the CRE measure that related and based on the image. Secondly, the optimal solution of CRE function found. Finally, the proposed method is applied over data set of image such as nondestructive testing (NDT) images. Moreover, we compare this with several classic segmentation techniques on the same data set.