Abstract:

Two-dimensional cross entropy thresholding method proposed recently is based on a gray level average gray level histogram which is wrongly divided. Although the recursive algorithm is adopted, the whole search space still has to be traversed for the optimal threshold, and the running speed needs to be further improved. Thus, an improved two-dimensional gray level-gradient histogram is given. The corresponding formulas of threshold selection based on two-dimensional minimum cross entropy and its recursive -algorithm are derived. And the chaotic particle swarm optimization (PSO) algorithm based on the improved Tent map is used to search for the two-dimensional optimal threshold, so as to reduce the running time. A large number of experimental results and a comparison with the existing two-dimensional cross entropy method based on gray level-average gray level histogram show that the proposed method takes almost all the object points and background points into account while computing the optimal threshold. As a result, it makes the segmentation results more accurate. Meanwhile, only a small part of the solution space needs to be searched to find the optimal threshold, and the required running time reduces to about 10%~40% of the original level.