为了有效地滤除红外图像中的噪声,提出了一种改进型非局部均值滤波算法(INLMF).该算法首先针对传统算法中采用固定尺寸的方形图像块无法有效刻画图像中大量分布的细节信息这一缺陷,结合图像中像素点灰度信息提出了一种图像块自适应划分方法,使得划分后的图像块在尺寸和形状上依赖于图像中灰度信息的实际分布情况;其次引入结构相似度因子对算法中的图像块权重值计算方法进行适当改进;最后分别将INLMF算法与已有的两类改进型NLMF算法对两幅红外监控图像进行滤波,并进行了理论分析和实验验证.结果表明,INLMF算法相对于其余几类算法而言,滤波效果较好,该研究对于提高红外图像滤波效果是有帮助的.

In order to filter the noise in infrared images effectively,an improved non-local mean filtering (INLMF) algorithm was proposed.In the traditional non-local means filtering (NLMF) algorithm,the square image blocks of fixed size cannot depict the image details effectively.For overcoming the defects of NLMF,a novel adaptive classification method of image blocks,combing with gray scale information of image pixels,was put forward.The divided image block in size and shape depended on the actual distribution of gray scale information.And then,structure similarity (SSIM) factor was introduced to improve the calculation method of image blocks weights.Two infrared monitoring images were filtered by two traditional NLMF algorithms and the new INLMF algorithm.The theoretical and experimental results show that the performance of INLMF is superior to the others.It is helpful for enhancing the filtering effects of infrared images.