联合多尺度块匹配的非局部均值去噪算法

针对非局部均值（Non-Local Means， NLM）图像去噪算法易产生伪影与平滑细节的问题，提出一种联合多尺度图像块匹配的像素相似性测度，提高NLM算法去噪性能。首先，研究与分析了加权欧氏距离与欧氏距离两种相似性度量以及图像块尺寸设置对NLM算法的影响。其次，通过引入图像特征信息并利用K-means聚类方法将图像划分为平坦区域和包含边缘与纹理的结构区域，对每个类别中的像素点，联合两种尺度图像块匹配计算像素的平滑权重。最后，优化了算法的滤波参数。实验结果表明，提出的算法在噪声去除与细节保持方面明显优于经典的NLM算法，相比其他改进的NLM算法也有优势。

Non-local Means Denoising Algorithm Derived from Combined Multi-scale Block Matching

Aiming at the problems that Non-Local Means (NLM) algorithm for image denoising tends to produce artifacts and smooth details, in this paper, multi-scale matching combination of image blocks was adopted to measure the similarity between pixels, which can improve the denoising performance of NLM algorithm. First, two similarity metrics (weighted Euclidean distance and Euclidean distance) and image block size used in NLM were studied and analyzed. Secondly, the whole image was partitioned into flat region and structural region by introducing its feature information and using K-means clustering method. For pixels in each category, the smooth weights were calculated by combining the matching of two image blocks in different sizes. Finally, the optimal choice for filtering parameter was given. Experimental results show that the proposed method outperformed the classical NLM algorithm in terms of noise removal and detail preservation and also has advantages over other improved NLM algorithms.