Thresholding for Remote Sensing Images of Rivers Based on a Fast Iterative Algorithm of Two-Dimensional Tsallis Cross Entropy

To further improve the accuracy and speed of remote sensing image segmentation of rivers, a segmentation method for remote sensing images of rivers is proposed, based on a fast iterative algorithm of two-dimensional Tsallis cross entropy. Firstly, a fast iterative algorithm for threshold selection using one-dimensional Tsallis cross entropy is proposed since the existing Tsallis cross entropy thresholding is not computationally efficient enough. Then the two-dimensional Tsallis cross entropy threshold selection formulae based on gray level-neighborhood average gray level histogram are derived to further improve the segmentation accuracy. In addition, recursive algorithms are adopted to calculate the intermediate variables involved in criterion function to avoid their repetitive computation. As a result, the calculating speed is improved. Finally, a fast iterative algorithm for threshold selection using two-dimensional Tsallis cross entropy is proposed, and the corresponding algorithmic formulae are derived. Thus the amount of calculation is greatly reduced. A large number of experimental results show that, compared with four recent thresholding methods, the proposed method has obvious advantages in segmentation results for remote sensing images of rivers and algorithmic running time. It is a fast and effective segmentation method which can be used in river detection and classification system.