基于交叉熵极小化的图像边缘检测算法  

Edge Detection Algorithm Based on Cross-Entropy Minimization

作　　者：金永镐[1] 崔荣一[1] 金小峰[1] 

机构地区：[1]延边大学工程学院,吉林延吉133002

出　　处：《吉林大学学报（信息科学版）》2006年第6期656-660,共5页Journal of Jilin University（Information Science Edition）

基　　金：国家自然科学基金资助项目(69362001)

摘　　要：提出了一种自然图像边缘检测的算法,通过差分运算把边缘检测问题转化为目标检测问题,利用交叉熵极小化方法使背景区域和目标区域内的灰度级分布与差分图像灰度级的先验概率(广义高斯分布)相匹配,确定了背景与目标之间灰度分界阈值,检测出图像的边缘成分。实验表明,提出的方法具有良好的边缘检测性能。An edge detection algorithm is proposed in this paper for natural image. The problem of edge detection is converted to the problem of object detection by difference operation. The background-object gray-level threshold is determined by minimizing the cross-entropy between gray-level distributions of background and object and the a priori generalized Gaussian distribution of difference image. The edge component of image is detected by segmenting the object in difference image. The performance of edge diction of proposed algorithm is validated by experiment results.

关 键 词：边缘检测 差分图像 广义高斯分布 交叉熵 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]