Two-dimensional minimum local cross-entropy thresholding based on co-occurrence matrix

This paper introduces a novel image segmentation method that performs histogram thresholding on an image with consideration to spatial information. The spatial information is the joint gray level values of the pixel to be segmented and its neighboring pixels that are based on the gray level co-occurrence matrix (GLCM). The new method was obtained by extending the one-dimensional (1D) cross-entropy thresholding method to a two-dimensional (2D) one in the GLCM. Firstly, the 2D local cross-entropy is defined at the local quadrants of the GLCM. Then, the 2D local cross-entropy is used to perform the optimal threshold selection by minimizing. Results from segmenting the real-world images demonstrate that the new method is capable of achieving better results when compared with 1D cross-entropy and other classical GLCM based thresholding methods.

Figure optionsDownload as PowerPoint slideHighlights
► Gray level co-occurrence matrix can improves the image segmentation performance.
► Image local cross-entropies are defined based on gray level co-occurrence matrix.
► Optimal threshold is obtained by minimizing local cross-entropy.
► Proposed method makes higher values of uniformity and shape of segmented image.