一种基于散布矩阵的自适应非局部均值滤波模型

针对非局部均值滤波模型(non-local means model,NLM)中恒定的衰减系数不能同时实现图像不同特性结构区域去噪性能最优的缺点,提出了一种基于图像散布矩阵的自适应非局部均值滤波模型(scatter matrix non-local means model,SM-NLM).该模型构造图像的散布矩阵,通过散布矩阵的特征值确定衰减系数的大小,平滑区域采用较大的衰减系数,强纹理区域采用较小的衰减系数,以实现衰减系数自适应的非局部均值滤波.实验结果表明,本模型能取得更高的峰值信噪比,更好地保留了图像的细节信息.

A non-local means image denoising algorithm based on scatter matrix

In order to overcome that the decay parameter of non-local means model is constant and can''t reflect image''s local structural information, an adaptive non-local means model based on the scatter matrix is proposed. Firstly, scatter matrix of the image is constructed, then we use the scatter matrix''s eigenvalues to control the decay parameter, a relatively small decay parameter is chosen for an edge pixel while smooth region uses larger decay parameter, so different texture structure areas adopt different local decay parameter. Numerical experiments results show that this proposed adaptive model has higher peak signal to noise ratio, preserving more image details.