提出了一种基于梯度信息的结构相似性算法改进的红外图像非局部均值滤波方法。传统的非局部均值滤波算法采用欧氏距离度量图像块之间的相似性，因而不能够很好地衡量图像细节和边缘信息，导致滤波后图像模糊失真。针对此问题，采用结构相似性度量（structural similarity，SSIM）算法对欧氏距离进行加权改进，针对普通的SSIM边缘信息评价能力的不足，提出了带有梯度信息的GSSIM算法，实验结果表明本方法在保持非局部均值（Non-Local Means，NLM）滤波算法去噪能力的同时还能够较好地保持图像的边缘和细节信息。

In this paper, an improved infrared image non-local means (NLM) filtering method is proposed based on gradient information. The traditional nonlocal average filtering algorithm uses Euclidean distance to measure the similarity between image blocks. Therefore, it cannot measure the image detail and edge information accurately and causes blurring distortion after filtering. The structural similarity (SSIM) algorithm utilizes the weighted Euclidean distance. To have a better edge information evaluation ability, the GSSIM algorithm is used here with gradient information. The experimental results show that while maintaining the NLM filter denoising ability, this method is better able to preserve the image edge and detail information.