Image segmentation based on gray stretch and threshold algorithm

When the gray levels between the foreground and background do not change significantly in the image, the Otsu's method cannot determine the segmentation threshold accurately. In this paper, an image binarization method based on wavelet domain gray stretch is proposed. In order to emphasize the gray stretch and edge enhancements on the image, it combines the discrete wavelet transform with the principle of the largest between-class variance. Experimental comparisons between the method and other methods show that the proposed approach has a satisfactory image segmentation and strong robustness to noise.