构建了近地背景下的小目标图像模型, 提出了一种基于暗图像处理和改进非局部均值滤波的红外图像预处理算法。采用图像去雾思想, 将红外图像类比为有雾图像, 通过暗通道去雾方法实现近地背景抑制, 并引入相似度权值阈值改进双边 NLM(Non-Local Means)算法, 减少邻域内相似度权值较低的像素点对灰度值估计的影响, 得到预处理后的红外图像。实验表明, 对于近地背景下红外弱小目标检测, 该算法能够有效抑制背景, 增强目标与背景对比度, 提高红外图像预处理性能。

An infrared image model with small targets in the near-ground background is constructed. In addition, an infrared image preprocessing algorithm based on dark-image processing and improved non-local mean filtering is proposed. Using the theory of image dehazing, an infrared image is analogized as a foggy image, and near-ground background suppression is then achieved by the prior dark-channel method. In addition, similarity with a threshold is introduced to improve the bilateral non-local means algorithm; this reduces the influence of pixels in the neighborhood with low similarity on the estimation of the gray value and obtains the infrared image after preprocessing. The experiments show that the proposed algorithm can effectively suppress the background, enhance the contrast between the targets and the background, and improve the performance of infrared image preprocessing.