改进的最大熵阈值分割及其快速实现

Improved Two-dimensional Maximum Entropy Image Thresholding and its Fast Recursive Realization

针对传统二维最大熵阈值法对二维直方图采用近似处理等的不足,提出了改进的二维最大熵快速阈值分割方法。首先对邻域模板进行改进,将改进后的模板用来构建二维直方图,并将最大熵法用于此直方图上,以便获得最佳阈值;然后,舍弃传统的二维直方图中关于主对角区域的概率近似为1的假设,使阈值选取更准确;最后,分析二维直方图投影,得到其特性,并证明两定理的存在。利用此特性和两定理导出新型、快速的递推算法来降低计算复杂度。仿真实验结果表明,与当前二维最大熵法相比,提出的方法不仅分割更准确和抗噪性更强,而且占用的存储空间更少,分割速度更快,分割时间少于0.04s。

The traditional two-dimensional（2-D） maximum entropy（ME） thresholding method has not good segmentation performance mainly owing to approximately processing.So a fast and improved 2-D ME image thresholding method was presented in this paper.Firstly,a 2-D histogram with the improved neighborhood mask was given and the ME method was used on the 2-D histogram to get a more ideal threshold.Then,some values of objects area and background area in the 2-D histogram main-diagonal district in the ME method were calculated precisely to obtain better segmentation performance.Finally,a 2-D histogram was analyzed to get its features and two theorems were proved,and the features and the theorems were employed to infer a new recursive approach to search the best threshold vector to reduce the computational complexity.Experimental results show that the proposed method not only achieves more accurate segmentation results and more robust anti-noise,but also requires much less memory space and its running time is much less,around 0.04 second,compared to the current 2-D ME thresholding methods.