关键词: 阈值分割, 三维直方图, 指数灰度熵, 降维处理, 搜索策略

Abstract: In view of the existing threshold segmentation methods which usually only consider the statistical information from image histogram, while ignoring the gray distribution uniformity of the image target class and the background class, one-dimensional exponential gray entropy segmentation algorithm is put forward and three-dimensional exponential gray entropy segmentation algorithm is deduced. The principles of one-dimensional exponential gray entropy algorithm and three-dimensional exponential gray entropy algorithm are presented. The optimum segmentation threshold is got by combining dimension reduction and optimal search strategy on the three-dimensional histogram. The search complexity is reduced from [O(L3)] to [O(L1/2)] in theory. Experimental results show that, compared with other existing threshold algorithms, the proposed algorithm has better anti-noise performance, segmentation effect and its operation time is reduced greatly.

Key words: threshold segmentation, three-dimensional histogram, exponential gray entropy, dimension reduction method, search strategy