基于二维直分与斜分灰度熵的图像阈值选取

Gray Entropy Image Thresholding Based on 2-Dimensional Histogram Vertical and Oblique Segmentation

二维最大Shannon熵阈值选取方法仅依赖于图像二维直方图的概率信息,而没有直接考虑类内灰度的均匀性,为此本文提出了二维灰度熵的阈值选取方法.首先给出了灰度熵的定义及其一维阈值选取方法,该灰度熵与现有的仅基于直方图分布的最大Shannon熵不同,直接反映了类内灰度的均匀性;然后提出基于混沌粒子群优化的二维直分灰度熵阈值选取方法及其快速递推算法;最后导出了二维斜分灰度熵的阈值选取公式及其快速递推算法.实验表明,与基于粒子群优化的二维直分最大Shannon熵阈值选取方法、二维斜分最大Shannon熵阈值选取方法及二维斜分Otsu阈值选取方法相比,所提出方法的分割图像更能反映原始图像的边缘、纹理及细节信息.

The maximal Shannon entropy thresholding methods based on 2-dimensional histogram only depend on the probability information from 2-dimensional histogram of image, without considering the uniformity of within-cluster gray scale, so gray entropy tbresholding methods based on 2-dimensional histogram were proposed. Firstly, gray entropy was defined and one-dimensional thresholding method was given. Different from the maximal Shannon entropy only based on histogram distribution, the gray entropy reflects the uniformity of within-cluster gray level im- mediately. Then one-dimensional gray entropy thresholding formula was extended and the thresholding method based on 2-dimensional histogram vertical segmentation was proposed. Its fast recurring algorithm was given and chaotic particle swarm optimization was used to search for the best thresholds. Finally, the gray entropy thresholding method based on 2-dimensional histogram oblique segmentation was proposed, and its fast recurring algorithm was given. Experiments show that segmented images of the proposed methods can better reflect the edge, texture and details of the original image, compared with the maximal Shannon entropy thresholding method based on 2-dimensional histo- gram using particle swarm optimization, the maximal Shannon entropy thresholding method based on 2-dimensional histogram oblique segmentation and the Otsu thresholding method based on 2-dimensional histogram oblique segmentation.