基于邻域多样本系数估计的小波图像去噪  

Wavelet Image Eenoising Based on Multi-sample Estimation Using Neighboring Coefficients

作　　者：游林儒[1] 谢俊斌[1] 

机构地区：[1]华南理工大学自动化科学与工程学院,广州510641

出　　处：《计算机与数字工程》2008年第3期100-102,共3页Computer & Digital Engineering

摘　　要：根据信号估计理论推导了利用邻域系数对中心系数进行多样本最大后验概率(MAP)估计的比例萎缩公式,并结合平稳小波变换提出一种低复杂度的图像去噪方法。首先用平稳小波变换得到冗余的小波系数,再根据图像边缘在每个点的邻域内选择能共同反映物体内部或边界的同类点作为多个样本,利用比例萎缩公式对小波细节系数进行估计。实验表明,和现有方法相比,该方法具有更高的信噪比和更宽的噪声适应范围,在有效去除噪声的同时清晰的保留了图像边缘。According to signal-estimation theory,a scale shrinking formula was deduced in which every wavelet coefficient was estimated from several neighboring coefficient samples by using maximum a posterior(MAP)estimator,combined with this formula and stationary wavelet transform(SWT),a low-complexity image denoising method was proposed.The redundant wavelet coefficients were accessed through SWT firstly,secondly some similar points which reflected the objects' content or edge were chosen in each point's neighborhood according to the image edge,finally each wavelet detail coefficient was estimated from these similar points using the formula presented.Experimental results show that the proposed method has higher Signal Noise Ratio(SNR)and wider range of noise tolerance than some classical methods,and performs better both in removing noise and preserving image edges.

关 键 词：平稳小波变换 图像去噪 贝叶斯估计 边缘检测 

分 类 号：TN911.73[电子电信—通信与信息系统]