基于图像分割和配准的红外焦平面阵列非均匀校正算法  

Image Segmentation and Registration-Based Algorithm for Nonuniformity Correction of Infrared Focal Plane Arrays

作　　者：徐田华[1] 张小军[1] 赵亦工[1] 

机构地区：[1]西安电子科技大学模式识别与智能控制研究所,陕西西安710071

出　　处：《红外技术》2004年第3期5-10,共6页Infrared Technology

基　　金：国家自然科学基金(编号69982008)资助项目; 国家教委资助优秀年轻教师基金(编号:2000年度)资助项目。

摘　　要：在目标成像跟踪和制导领域,存在大量由缓慢移动背景和相对背景运动的点目标构成的红外图像序列,传统的基于场景的红外焦平面非均匀校正算法在解决此类图像中存在困难。提出了基于图像分割和配准的红外焦平面非均匀校正算法(简称S-R算法),通过将图像背景和运动点目标分离,利用图像配准的方法完成图像背景的非均匀校正和点目标位置处错误固定噪声参数的补偿,最终完成整个焦平面探测单元固定噪声参数的估计,从而有效解决了这类红外图像序列的非均匀校正问题。S-R算法具有噪声参数估计精度高、收敛速度快和计算复杂度低等优点。文中最后用仿真数据对上述结论进行验证。There are a lot of infrared image sequences consisting of slowly moving background and point object with relative motion t in the field of target imaging, tracing and navigation, but the traditional scene-based algorithm for nonuniformity correction (NUC) of infrared focal planes arrays (IRFPA) had difficulty in solving these types of images. In the paper, a image segmentation and registration-based algorithm (S-R algorithm) had been proposed. Through separating the background and the point object, the former was corrected by the registration-based algorithm and the error fixed-pattern noise (FPN) parameters in the point object was compensated. So the FPN in the IRPFA was estimated completely and the NUC was solved efficiently. S-R algorithm has the advantage of accurate estimation of FPN, fast convergence rate and low computational complexity. At the end the conclusions were verified by the simulation data.

关 键 词：图像分割 图像配准 红外焦平面阵列 非均匀校正 固定噪声 校正算法 评价指标 

分 类 号：TN215[电子电信—物理电子学]