基于目标与背景相异位移配准的非均匀校正算法  

Non-uniformity correction algorithm based on registration of different motions of target and background

作　　者：杨硕[1] 赵保军[1] 毛二可[1] 唐林波[1] 

机构地区：[1]北京理工大学,北京100081

出　　处：《激光与红外》2013年第11期1286-1290,共5页Laser & Infrared

摘　　要：本文针对场景图像中目标和背景存在着不同运动的情况,提出了一种基于图像分离和配准技术的校正算法,新算法通过傅里叶变换相位相关方法,估计出目标与背景的两种运动位移,然后通过计算匹配像素点的平均绝对误差将目标和背景进行标识区分,建立关于目标和背景位移量的误差平方和方程组,求解得到亚像素位移量,改进了原有的代数校正方法,分别根据目标和背景的位移顺序校正,并通过剔除无匹配点的双线性插值方法对新算法进行完善。计算机模拟图像序列与实际红外图像序列的实验结果表明,该算法比目前常用的代数校正算法具有更好的适应性,具有较高的实用价值。A new Non-Uniformity Correction algorithm based on registration of different motions of target and back-ground is proposed, for the situation of different displacements of target and background in infrared image. The new method estimates two of the motions of target and background by shift theory of Fourier Transform algorithm and iden-tifies target and background by Mean Absolute Difference algorithm. Motions of sub-pixel by establishing equations of sum square error of displacements of target and background are obtained. The ordinary method of Algebraic Non-Uni- formity Correction algorithm is improved by correction according displacements of target and background, and the new algorithm is completed through Bilinear Interpolation method with excluding no matching pixels. Computer simulations and actual experiments＇ result demonstrate, comparing with the ordinary Algebraic Non-Uniformity Correction algo-rithm,the proposed algorithm has the superiority, adaptability and a high practical value.

关 键 词：相异位移 傅里叶变换 平均绝对误差 双线性插值 代数校正 

分 类 号：TN911.73[电子电信—通信与信息系统]