Robust multilevel image thresholding of local 3×3 max grayscale and 2D entropy

The utilization of 2D entropy will enable the auto-segmentation for generating a 2D histogram. Conventional methods focus on the central grayscale and the mean grayscale of pixels within the same neighborhood. However, in reality, the horizontal or vertical variation in grayscale fails to comprehensively represent the overall characteristics among neighboring pixels. To this end, a maximum grayscale approach is put forward. This method constructs a 2D histogram by the central grayscale and the maximum grayscale among the grayscales of all pixels within the same 3×3 neighborhood. Extensive experiments on nearly all the images in the Matlab image toolbox demonstrate the efficacy of this novel method.