Gray image thresholding based on three-dimensional Renyi entropy

Most of the current thresholding algorithms for gray images conduct segmentation on the 2-dimensional histogram, resulting in the low precisions for details and the weak immunities for noises. A novel thresholding algorithm based on 3-dimensional Renyi´s entropy is proposed in this paper. The new dimension added is the weighted neighbor median grayscale, which has shown a high effectiveness on noise removal and detail reservation, thereby improved the segmentation quality. The experiments on different images indicated that our algorithm has increased the precision and raised the noise immunity compared with three types of 2-dimensional thresholding methods.