摘要： 经典非局部均值去噪算法在度量邻域结构特征相似性上对于旋转和平移情况下不够灵活。为此,提出一种基于归一化转动惯量的非局部均值图像去噪算法。利用归一化转动惯量作为相似权重的一部分来调节相似性度量。实验结果表明,与原始非局部均值去噪算法相比,该算法能够更好地保持图像的细节结构信息,提高峰值信噪比和结构相似度,特别当噪声强度增大时,峰值信噪比最大提高1.03 dB,结构相似度最大提高0.04。

关键词: 非局部均值, 去噪, 相似性度量, 噪声图像, 归一化

Abstract: In order to solve inflexible rotation and translation in measuring neighborhood similarity for structural characteristics in Non-local Means(NLM) denoising method,this paper presents the NLM image denoising algorithm based on Normalized Moment of Inertia(NMI).The normalized moment of inertia as a part of the similar weight is proposed to balance neighborhood patches’ similarity.Experimental results show that compared with the original nonlocal mean denoising algorithm,this algorithm can keep the detail of the image structure information preferably,and can improve the Peak Signal to Noise Ratio(PSNR) and structural similarity.Especially,when the noise intensity increases,the algorithm improves the PSNR by 1.03 dB and the structure similarity index by 0.04.

Key words: Non-local Means(NLM), denoising, similarity measurement, noise image, normalization

中图分类号: 

• TP391