基于引导核聚类的非局部均值图像去噪算法

Nonlocal Means Image Denoising Algorithm Based on Steering Kernel Clustering

为改善非局部均值(NLM)算法对不规则纹理图像的去噪效果,提出了一种基于引导核聚类和自适应搜索窗的NLM图像去噪算法。首先使用基于引导核的模糊C均值(FCM)聚类算法对相似窗进行预筛选,划分其类别;然后根据相似窗的类别计算每个像素点对应的搜索窗大小,保证相似性较高的相似窗数量;最后分别对每一类进行自适应搜索窗的NLM图像去噪。实验结果表明:与基于Zernike矩、基于主邻域字典(PND)、基于均值方差预筛选等3种NLM改进算法相比,该NLM改进算法对强噪声污染或不规则纹理的图像,其去噪效果更为有效,并更好地保持了图像的纹理、边缘,在峰值信噪比(PSNR)和结构相似性测度(SSIM)等客观定量评价指标上优于其他NLM改进算法。

In order to improve the denoising effect of nonlocal means （NLM） algorithm for irregular texture images, an image denoising algorithm of NLM based on clustering by steering kernel and adaptive search windows is proposed in this paper. Firstly, fuzzy c-means （FCM） clustering algorithm based on steering kernel is used to prescreen and classify similar windows. Then, the size of search windows corresponding to each pixel is calculated according to categories of similar windows. The number of similar windows with higher similarity is guaranteed. Finally, image denoising of NLM based on adaptive search windows is carried out for each category. A large number of experimental results show that the proposed improved NLM algorithm has better denoising effect for the images with strong noise or irregular texture images, compared with the three improved NLM algorithms which are based on Zernike moment, principal neighborhood dictionaries （PND）, and prescreening of mean-variance, respectively. The textures and edges in images are better preserved. The proposed algorithm is superior to other improved NLM algorithms in objective quantitative evaluation indexes such as peak signal to noise ratio （PSNR） and structural similarity index measurement （SSIM）.