基于数学形态学变形虫的自适应图像滤波  

Adaptive Image Filter Algorithm Based on the Mathematical Morphology Amoebas

作　　者：王艳华[1,2] 刘伟宁[1] 陈爱华[1,2] 

机构地区：[1]中国科学院长春光学精密机械与物理研究所,吉林长春130033 [2]中国科学院研究生院,北京100039

出　　处：《激光与红外》2008年第12期1258-1261,共4页Laser & Infrared

基　　金：中国科学院二期创新基金项目(No.C04708Z)资助

摘　　要：针对形态学滤波中的一个关键问题——结构元的选择,提出了一种基于数学形态学的变形虫结构元,它是一种没有固定形状和大小的结构算子。按照导引图像建立的梯度准则,以像素为中心在其邻域内进行结构元的生长,然后在每个像素点利用所生成的结构元进行灰度形态学中值或者均值滤波。实验结果表明:这种非线性滤波算法能够很好地去除噪声、增强边缘,克服了传统滤波的缺点,而且相对于经典保边滤波各向异性扩散法具有计算量小的特点。In view of the selection of structure elements problem in morphological filter, this paper presents morphological operators with non-fixed shape and non-fixed size kernels, or amoebas, instead of fixed-shape, fixed size structuring elements generally used. These structure kernels take into account the image contour variations to adapt their shape and size according to the gradient criteria built firstly. Then, it performs the median or mean filtering at each pixel with these structure kernels. Experiments results show that this method can simultaneously suppress the clutter and enhance edges to overcome the transitional filter＇s drawbacks, simultaneously comparing the anisotropic filter has better efficiency.

关 键 词：非线性滤波 自适应邻域形态学 结构元 扩散滤波 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]