基于图像配准的非均匀性校正算法  

Nonuniformity correction algorithm based on image registration

作　　者：李成立 吕俊伟 刘亮 王佩飞 

机构地区：[1]海军航空大学,山东烟台264001

出　　处：《激光与红外》2018年第4期531-536,共6页Laser & Infrared

基　　金：国家自然科学基金项目(No.61205206)资助

摘　　要：非均匀性严重影响红外制导武器的成像质量和系统性能,有效进行非均性校正是提高红外制导武器成像质量的有效手段。高通滤波类算法对非均匀性校正的参数漂移具有很强的适应性,但易产生"鬼影"现象,影响校正结果。本文针对高通滤波类算法的缺陷,结合帧间配准技术,提出了一种基于图像配准的非均匀性校正算法。与SLTH算法和BFTH算法中利用滤波器滤波后得到的残差图像相比,本文利用图像配准和场景删除得到的残差图像能更好地消除图像中的场景信息,抑制"鬼影"效果显著。实验表明:本文算法的校正效果良好,校正结果的平均粗糙度低至3.79,优于SLTH和BFTH两种算法。而且本文算法处理速度快,达到了每秒15.83帧,是BFTH处理速度的1.63倍。Nonuniformity has great influence on the image quality and system performance of infrared guided weapons,so good nonuniformity correction is effective means to improve imaging quality of infrared guided weapons.High-pass filtering algorithm has a strong adaptability to parameter drift of nonuniformity correction,but it is easy to produce the“ghost”images,which degrades the correction results.Aiming at the defects of high-pass filtering algorithm,a new nonuniformity correction algorithm based on image registration is proposed in combination with interframe registration technology.Compared with the residual image obtained from the filter in SLTH algorithm and BFTH algorithm,the residual image obtained by the proposed method can better eliminate the scene information of the image and can restrain the“ghost”effect markedly.The experimental results show that the correction effect of this method is good,the average roughness of correction results is as low as 3.79,which is much lower than SLTH and BFTH algorithm.Besides,the processing speed of the algorithm is fast,and it reaches 15.83f/s and is 1.63 times the speed of BFTH processing.

关 键 词：红外成像系统 非均匀性校正 图像配准 鬼影现象 

分 类 号：TN215[电子电信—物理电子学] TP391.41[自动化与计算机技术—计算机应用技术]