基于模糊指数熵和模拟退火的图像分割

Image Segmentation Based on Fuzzy Exponential Entropy and Simulated Annealing Algorithm

为了有效分割照度不均匀的网格图像,提出了一种基于模糊指数熵和模拟退火算法的阈值分割方法。基于模糊集合理论,根据像素灰度值把原始图像中的像素分为黑和亮两个模糊集,利用最大模糊熵准则确定模糊区间的范围,寻找模糊参数的最优组合,确定最优分割阈值,实现图像分割。由于使用穷举法搜索模糊参数的最优组合存在计算复杂度高、占用存储空间大的弱点,因此采用模拟退火算法确定最优阈值,从而减少了计算量。实验结果表明,此方法能够自动、有效地选取阈值,运算时间约为使用穷举法寻求最优阈值所需时间的1/3,并且分割效果明显优于最大类间方差法、迭代法和一维最大熵法。

A thresholding method based on fuzzy exponential entropy and simulated annealing algorithm is proposed to segment a gridding image with nonuniform illumination efficiently. Based on the fuzzy set theory, the method can classify the pixels of an image into two parts, namely, dark and bright part by their gray levels, define the fuzzy region and search the optimal combination of the fuzzy parameters by the maximum fuzzy entropy principle. But it has a high computational complexity and occupies large memory size for an exhaustive search. A simulated annealing algorithm is implemented to search the optimal combination of the fuzzy parameters, which finally determine the threshold, and reduce computational load. Experimental results show that the proposed method can select the threshold automatically and efficiently, and its computational load is about one third that of an exhaustive search. It is quite evident that the proposed method has an advantage of segmentation result over the Otsu＇s method, the iteration method and the one-dimensional maximum entropy method.