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.
出处
《电光与控制》
北大核心
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)
关键词
合成孔径雷达
图像检测
多重分形
奇异谱
边缘检测
synthetic aperture radar
image detection
multiple fractal
singularity spectrum
edge detection
