广义模糊熵的构造及其在图像分割中的应用

Construction of generalized fuzzy entropy and its application in image segmentation

针对现有的模糊熵阈值法对噪声干扰、光照不均匀图片的分割不能获得满意分割效果的不足,提出了模糊熵公式的参数化修改并用于图像分割,获得了选取最佳分割阈值的模糊熵新算法。首先将Sugeno广义否定算子替代二次型模糊熵公式的Zadeh标准否定算子,得到了参数型的广义模糊熵表达式;其次,将其应用于构造图像阈值化分割的准则函数;最后,给出了图像阈值化分割新算法中的广义模糊熵参数自动选取方法。实验结果表明,给出的广义模糊熵图像分割方法对光照不均匀图像相比传统模糊熵分割方法更有效。

To deal with the shortcoming of the classical fuzzy entropy thresholding method that it is incapable of gaining satisfactory segmentation quality of the image with noise or uneven illumination, a new image thresholding segmentation algorithm based on fuzzy entropy modified in parametric form is proposed. Firstly the fuzzy entropy in quadratic form is modified by substituting standard negative operator of Zadeh with generalized negative operator of Sugeno, and the generalized fuzzy entropy formula is obtained. Then the criteria function of image thresholding segmentation is constructed based on the generalized fuzzy entropy in parametric form. Lastly, the method of automatic parameter selection of generalized fuzzy entropy applied in image segmentation is proposed. Experimental results show that the new segmentation algorithm based on the generalized fuzzy entropy can get better vision quality on the images with uneven illumination than that of classical fuzzy entropy.