基于变尺寸窗口的多特征NLM图像去噪算法  

Multi-features NLM Algorithms Based on Variable Size Window for Image De-noising

作　　者：毛静 MAO Jing(College of Electronic and Information Engineering,Ankang University,Ankang 725000)

机构地区：[1]安康学院电子与信息工程学院,安康725000

出　　处：《计算机与数字工程》2023年第5期1138-1143,1149,共7页Computer & Digital Engineering

基　　金：国家自然科学基金面上项目(编号:12174004);安康市科技计划项目(编号:AK2020-GY03-2)资助。

摘　　要：常规非局部均值算法的邻域相似性计算过程容易遭受噪声干扰,影响相似像素的权重分配,导致图像结构信息损失严重。针对上述问题,提出一种变尺寸窗口的多特征非局部均值算法,首先根据图像结构张量特征对图像区域进行划分,在不同特征的区域内采用不同尺寸搜索窗口和自适应平滑滤波参数,结合灰度特征、多方向梯度特征和空间特征共同度量邻域相似性,再应用双核函数计算相似性权值,对目标区域的像素权值进行重分配,从而实现图像去噪目的。结果表明,新改进方法的图像峰值信噪比平均提高69%以上,结构相似度平均达到0.77以上。结论认为,相比常规非局部均值算法,新改进方法去噪能力强,边缘及纹理细节保护更好,具有良好的应用前景。The calculation process of the similarity between neighbors with the conventional non-local mean algorithm is susceptible to noise interference,which affects the weight distribution of similar pixels,resulting in serious loss of image structure information.In response to the above problems,a multi-feature non-local mean algorithm with variable-size windows is proposed.First,the image area is divided according to the tensor characteristics of the image structure,and different size search windows and adaptive smoothing filter parameters are used in the areas with different characteristics.The gray-scale feature,multi-directional gradient feature and spatial feature jointly measure the similarity of the neighborhood,and then the dual kernel function is used to calculate the similarity weight,and the pixel weight of the target area is redistributed,so as to achieve the purpose of image de-noising.The results show that the image peak signal-to-noise ratio of the new improved method increases by more than 69%on average,and the structural similarity reaches an average of more than 0.77.The conclusion is that compared with the conventional non-local mean algorithm,the new improved method has strong de-noising ability,better protection of edge and texture details,and has good application prospects.

关 键 词：非局部均值 多特征 双核函数 结构张量 搜索窗口 信噪比 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]