灰度图像最大熵分割方法的改进

Improvement of the Gray Image Maximum Entropy Segmentation Method

传统的最大熵分割方法只考虑了图像的灰度概率,忽略了对应的灰度值。为了充分利用灰度图像的灰度信息和空间信息,改进了传统的二维灰度直方图,生成二维差值属性灰度直方图。另外,改进了灰度均值和二维熵的计算方法。在计算熵时,以二维差值属性灰度直方图为基础,用空间信息值来代替灰度概率,生成二维差值属性信息值熵。在实验中,对多张不同的灰度图像分别用改进的最大熵方法与传统的最大熵分割方法进行分割,并对分割结果进行比较分析。实验结果表明,改进的最大熵分割方法能有效地分割灰度图像及噪声图像,有很强的抗噪声能力,并能产生清晰的分割结果。

In the traditional maximum entropy threshold segmentation methods,the gray probability of image is used and the corresponding gray value is ignored.In order to adequately utilize the gray information and spatial information of the gray image,the traditional 2D gray histogram is improved and the 2D D-value attribute gray histogram is formed.Otherwise,the computation method for the average gray value and the 2D entropy is improved.We use the spatial information value as a substitute for the gray probability to compute the entropy.The computation of the entropy is based on the D-value attribute gray histogram and creates the spatial different attribute information value entropy（SDAIVE）.In experiments,many different gray images are segmented with the improved maximum entropy method and the traditional maximum entropy method,and the segmentation results are compared.The experimental results show that improved threshold method can effectively segment gray images and noise images.This method has strong anti-noise capability and clear segmentation results.