基于图像方差和信噪比的红外弱小目标检测算法  

Infrared Weak and Small Target Detection Algorithm Based on Image Variance and Signal-To-Noise Ratio

作　　者：韩郡郡 马骏 王博凯 HAN Junjun;MA Jun;WANG Bokai(College of Ordnance Science and Technology,Xi'an Technological University,Xi'an 710021,China)

机构地区：[1]西安工业大学兵器科学与技术学院,陕西西安710021

出　　处：《红外技术》2024年第11期1293-1301,共9页Infrared Technology

摘　　要：针对复杂场景下存在背景变化剧烈、杂波较多、信噪比低而造成红外弱小目标检测准确率低、虚警率高的问题,提出利用图像全局信息与局部对比度结合的红外弱小目标检测算法。算法利用方差和信杂比对图像的所有像素点统计分析,将图像做全局处理得到特征图,以适应灰度变化剧烈的复杂背景,同时抑制大量平缓背景杂波,提高目标的信杂比。针对特征图中主要存在的强背景边缘噪声和高亮像素点噪声,采用加权绝对方向平均差(weighted absolute directional mean difference,WADMD)算法,将目标与背景的强度差异作为加权系数,计算目标与各方向背景的绝对平均差,并使用判断门限抑制负对比度,抑制高亮度噪声的同时,提高目标的显著性。实验表明,与对比算法相比,本文算法能够适应多变的复杂背景,对目标信杂比提升更明显,鲁棒性更好。Aiming at the problems of low detection accuracy and high false alarm rate of infrared weak and small targets due to violent background changes,more clutter and low signal-to-noise ratio in complex scenes,an infrared weak target detection algorithm combining image global information and local contrast is proposed.The algorithm uses the variance and signal-to-noise ratio to statistically analyze all pixels of the image,and processes the image globally to obtain a feature map,so as to adapt to the complex background with sharp gray level changes,while suppressing a large number of flat background clutter and improving the signal-to-noise ratio of the target.Aiming at the strong background edge noise and bright pixel noise mainly existing in the feature map,the weighted absolute directional mean difference(WADMD)algorithm is used to calculate the absolute average difference between the target and the background as the weighting coefficient,and the judgment threshold is used to suppress the negative contrast,and suppress the high luminance noise,and improve the significance of the target.Experiments show that compared with the comparison algorithm,the proposed algorithm can adapt to the changeable complex background,and improve the signal-to-noise ratio of the target more obviously,and have better robustness.

关 键 词：红外图像 目标检测 多尺度局部对比度 图像方差 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]