保持图像细节的多尺度形态滤波的新方法  

Novel Multiscale Morphological Filtering Method for Preserving the Details of Images

作　　者：杨照华[1] 浦昭邦[1] 祁振强[2] 

机构地区：[1]哈尔滨工业大学自动测试与控制系,黑龙江哈尔滨150001 [2]哈尔滨工业大学控制科学与工程系,黑龙江哈尔滨150001

出　　处：《光电子．激光》2003年第8期862-865,共4页Journal of Optoelectronics·Laser

基　　金：黑龙江省自然科学基金资助项目(F01 04);哈尔滨工业大学跨学科交叉性研究基金资助项目(HIT2000.83)

摘　　要：提出了一种适于灰度图像细节的多尺度形态滤波(MMFPD)方法。在普通多尺度形态开、闭滤波基础上,增加了多尺度top hat变换和bottom hat变换,用于提取并平滑小于该心渡的特征。此外,较小的尺度特征中包含噪声颗粒的可能性较大,据此修正了top hat变换和bottom hat变换的系数。对标准灰度图像进行滤波处理的实验结果表明,该方法噪声去除效果好,图像细节保持完整,处理效果明显优于传统滤波方法。A novel multiscale morphological filtering method for preserving the details of image(MMFPD) was presented by adding the multiscale tophat transformation and bottomhat transformation to the conventional multiscale morphoogical opening and closing filtering.The two added transforms were used to extract and smooth the features which are smaller than the current scale.It was found that the smaller features have greater possibility to contain noise particles.Accordingly,the coefficients of tophat transformation and bottomhat transformation were modified.Simulation results on the standard graylevel images show that the method can effectively reomve noise and preserve the details of images completely,and demonstrate better performance than the conventional filtering methods.

关 键 词：多尺度形态滤波 MMFPD TOP-HAT变换 bottom-hat变换 灰度图像 图像处理 数学形态学 

分 类 号：TN911.73[电子电信—通信与信息系统]