基于非局部均值的彩色图像去噪

摘要

Since the existing denoising methods of color images are usually hierarchical denoising of every channel,and they don’t consider the correlation of RGB channels,a non-local mean(NLM)denoising method of color image is proposed on the basis of similarity estimation of RGB channels. When non-local means filtering is used to remove the noise of color image, the target pixel is taken as the center to determine its support region,and then calculate the weight according to similarity esti-mation of RGB channels and perform the filtering of every channel by the one-by-one filtering method. A new noise parameter evaluation method is proposed to estimate the parameter of Gaussion noise in color image. Experimental results show that the al-gorithm proposed in this paper,compared with the traditional algorithms,has improved the PSNR and FSIM more,and it is necessary to consider the correlation among RGB channels in the process of denoising. The experimental results also proved the effectiveness of the new algorithm.%传统的彩色图像去噪算法通常是分层处理的，而忽略了彩色图像RGB通道之间的相关性，因此基于RGB通道联合相似度估计提出了一种新的彩色图像非局部均值去噪方法。在用非局部均值滤波对彩色图像进行去噪时，首先以目标像素为中心确定其支撑区域，然后根据多通道联合相似度估计确定权重，最后采用逐块滤波的方法对每一层进行滤波。并且针对彩色图像中含有的高斯噪声提出了一种新的噪声参数估计方法。由实验结果可以看出该算法比传统的去噪算法在PSNR和FSIM方面都有提高。因此可以看出在图像去噪过程中考虑三通道之间的相关性是必要的，同时也证明了算法的有效性。