基于直方图指数平滑的模糊散度图像分割

Histogram exponentially smoothed based image segmentation using fuzzy divergence

当图像灰度直方图具有多峰时,模糊散度图像分割方法的参数初始化有困难。针对这一问题,基于直方图指数平滑峰点自动检测原理,提出了一种自适应的模糊散度图像分割算法。该算法首先利用直方图指数平滑峰点自动检测原理,检测出直方图合理的波峰点和波谷点,并统计波峰的个数。然后利用检测到的波谷值初始化目标平均灰度等级搜索区间,利用波峰值初始化散度函数搜索区间,分类数由检测到的峰点个数来确定。实验结果表明,该方法更具有普遍性和适应性。

To overcome the difficulty of parameter initializing in fuzzy divergence image segmentation method when histogram have many peaks, an adaptive image segmentation algorithm is proposed based on peak searching of histogram exponentially smoothed. First, the peaks and valleys of histogram are detected using histogram exponentially smoothed approach and calculate the number of detected peaks. Second, the detected valleys are used to initialize object＇s average gray level searching intervals, the detected peaks are used to initialize divergence searching intervals and the class numbers is determined by the number of peaks. Experimental results show that the proposed algorithm is more universal and adaptive.