Improved Image Thresholding Based on 2-D Tsallis Entropy

The 2-D maximum Tsallis entropy (2DMTE) method not only considers the distribution of the gray information and the spatial neighbor information with using the 2-D histogram of the image, as a global threshold method, but also it often gets better segmentation results and flexibility owing to a parameter than other 2-D entropy methods. However, its performance is sensitive to its parameter. How to choose the parameter is often an obstacle in real time application systems. In this paper, improved image segmentation based on two-dimensional Tsallis entropy is presented. Firstly, the parameter of Tsallis entropypsilas method is changed into two parameters, then, the middle value of the image probability distribution is obtained, finally, according to it, the parameters are selected adaptively. Experimental results show the proposed approach can get much better segmentation result than the previous 2-D thresholding methods.