摘要： 在传统二维最大熵图像阈值分割算法中，二维直方图主对角区域的概率和近似为1的假设不够合理，且算法耗时较多。为此，提出一种新的最大熵分割算法。根据灰度级和韦伯局部描述子(WLD)建立二维WLD直方图(2D-WLDH)，将其用于最大熵的阈值分割，并设计快速递推算法，以提高运行速度。实验结果表明，该算法的运行时间较少，分割效果较好。

关键词: 图像分割, 阈值选取, 韦伯局部描述子, 最大熵, 二维直方图

Abstract: The traditional 2D maximum entropy threshold segmentation algorithm has an inadequately reasonable assumption that the sum of probabilities of main-diagonal distinct is approximately one in the 2D histogram and the algorithm is time-consuming. Aiming at this problem, a new maximum entropy segmentation algorithm is proposed in this paper. Based on gray level and Weber Local Descriptors(WLD), it constructs a 2D WLD Histogram(2D-WLDH), and applies it to the maximum entropy threshold segmentation. In order to further improve the speed of the proposed algorithm, the fast recursive algorithm is deduced. Experimental results show that, compared with existing corresponding algorithms, the proposed algorithm can reduce the running time and achieve better segmentation quality.

Key words: image segmentation, threshold selection, Weber Local Descriptor(WLD), maximum entropy, 2D histogram

中图分类号: 

• TP391