基于Clifford代数的多光谱图像边缘检测  

Edge detection of multispectral image based on Clifford algebra

作　　者：刘辉[1] 徐晨[1] 曹文明[1] 

机构地区：[1]深圳大学信息工程学院,深圳518000

出　　处：《东南大学学报（自然科学版）》2012年第2期244-248,共5页Journal of Southeast University：Natural Science Edition

基　　金：国家自然科学基金资助项目(61070087);深圳市基础基金资助项目(JC201005800305280570A;JC201105080030534011A)

摘　　要：为了充分利用多光谱图像不同图层之间的关联性,采用Clifford代数描述多光谱图像.在Clifford代数空间中,定义了多光谱图像的Clifford微分与Clifford梯度.在此基础上,提出了一种新的多光谱图像边缘检测与融合算法.该算法首先计算出各像素点的Clifford梯度,进而得到Clifford梯度范数;然后以此为依据,判断像素点是否为边界点,从而得到多幅边缘检测图像;最后,将这些图像融合,便可得到最终的边缘图像.与基于最大熵的多光谱图像边缘检测算法的比较结果表明,算法由于利用了多光谱图像不同图层之间的关联性,因而可以更好地保留边缘信息,获得更完整的边缘检测效果.In order to make a full use of the connection among different layers of a multispectral im- age, the Clifford algebra is used to describe the multispectral image. The Clifford derivative and the Clifford gradient of the multispectral image are defined in the Clifford algebra space. Then a new edge detection and fusion algorithm is proposed by analyzing the edge detection and fusion of the multispectral image. In the algorithm, the Clifford gradient for each pixel in the image is computed and the corresponding Clifford normal number is obtained. Whether a pixel is the edge pixel is then judged with the normal number, and several edge detection images are obtained. Finally, the last edge image is deduced by fusing these edge detection images. The comparison results between the proposed algorithm and the edge detection algorithm based on the maximal entropy show that by tak- ing full advantage of the connection among different layers of the multispectral image, the proposed algorithm can keep more integrated edge information of the multispectral image and obtain more complete edges.

关 键 词：多光谱图像 边缘检测 CLIFFORD代数 Clifford梯度 图像融合 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]