一种基于多重分形的SAR图像边缘检测方法  

SAR image segmentation based on multi-fractal

作　　者：张红蕾[1] 宋建社[1] 张宪伟[2] 

机构地区：[1]第二炮兵工程学院信息工程研究所,西安710025 [2]第二炮兵装备研究院信息所,北京100085

出　　处：《电光与控制》2007年第5期86-88,101,共4页Electronics Optics & Control

基　　金：国家自然科学基金资助项目(60272022)

摘　　要：分形维数只能刻画那些具有理想的自相似性的分形体,现实中的许多纹理并不满足这一条件,因此单一的分形维数并不足以描述和刻画SAR图像的纹理,多重分形维数更适合于描述图像的纹理。通过计算原始SAR图像离散点数据的奇异性指数,然后对应每一点奇异性指数计算全局多重分形奇异谱,根据判决准则区分边缘和纹理可以实现SAR图像的边缘检测,实验结果表明,基于多重分形特征的边缘检测算法能够检测到许多局部细节,同时又避免出现不重要的细节,突出了主要的边缘信息,很好地区分出SAR图像的纹理和边缘。The texture of SAR images can be described by fractal dimension, but single fractal dimension can not be enough, because fractal dimension can only describe those ideal self-similar fractal object, so multi-fractal is more suitable for describing the texture of images. The discrete image pixels＇ singularity index of SAR image and global multi-fractal singularity spectrum are calculated out based on multi-fractal principle, and edge extraction is realized by means of edge decision rule. The experimentation showed that this edge detection algorithm based on multi-fractal feature can detect many local details and avoid unessential details, also can distinguish texture and edges of SAR image.

关 键 词：合成孔径雷达 图像检测 多重分形 奇异谱 边缘检测 

分 类 号：V271.4[航空宇航科学与技术—飞行器设计]