摘要
为了有效地去除噪声,获得细节清晰的图像,提出一种噪声参数未知情况下基于小波分析的图像去噪算法.首先对图像上交互选定的不含细节的样本区域进行小波分解,根据噪声方差在不同尺度、不同方向子带内的映像建立噪声的特征向量;利用此特征向量计算小波分解后高频子带内各小波系数之间的相似度,把该相似度作为权值对当前小波系数进行调整.实验结果表明,文中算法对于噪声方差未知的图像不仅能有效地去除噪声,而且能保持图像的边缘信息,获得较好的去噪效果;对实拍的含噪数码照片进行测试的效果理想.
In this paper, we propose a non-local means algorithm for image denoising conducted in the wavelet domain with unknown noise variance. Firstly, a sample region containing little texture is selected by the user in the noisy image, then the noise variance is estimated by applying wavelet analysis, and a noise feature vector is constructed to characterize the noise of the entire image at different wavelet bands based on the estimation. For each high frequency sub-band, our proposed algorithm measures the similarity between each pair of wavelet coefficients and adjusts the value of each coefficient based on weighted mean of all the similar wavelet coefficients in the same sub-band. Experimental results demonstrate that our method can reduce noise effectively without blurring image features.
出处
《计算机辅助设计与图形学学报》
EI
CSCD
北大核心
2009年第4期526-532,共7页
Journal of Computer-Aided Design & Computer Graphics
基金
国家自然科学基金(60403038
60703084
60723003)
关键词
图像去噪
小波分析
非局部均值
image denoising
wavelet analysis
non-local means
