红外焦平面阵列非均匀性自适应校正算法中目标退化与伪像的消除方法  

ELIMINATING THE "GHOSTING" ARTIFACT AND TARGET FADE-OUT IN ADAPTIVE NONUNIFORMITY CORRECTION FOR IRFPA

作　　者：张天序[1] 石岩[1] 

机构地区：[1]华中科技大学图像识别与人工智能研究所

出　　处：《红外与毫米波学报》2005年第5期335-340,共6页Journal of Infrared and Millimeter Waves

基　　金：国家自然科学基金重点资助项目(60135020);国家重点预研项目(413010701-3)

摘　　要：深入分析了红外焦平面阵列(IRFPA)非均匀性的神经网络校正法出现目标退化和伪像的成因,指出没有考虑目标边缘而盲目更新系数是产生问题的根源.在此基础上提出了防止目标退化和伪像的边缘指导的神经网络自适应校正方法(ED-NN-NUC).仿真实验以及针对实际红外图像的实验结果表明,所指出的问题根源是正确的,提出的方法是合理有效的.The cause of the target fade-out and the ghosting artifact in NN nonuniformity correction scheme for infrared focal plane array （1RFPA） was studied. It was found that ignoring target edge and updating correction coefficients blindly in the NN-NUC scheme were the cause of the problems. Based on this consideration, an edge-directed NN-NUC scheme （ ED-NN-NUC ） was proposed to eliminate the ghosting artifact and the target fade-out. The comparison experiment with simulated data and real 1RFPA infrared data shows that our viewpoint is correct, and the proposed scheme is effective.

关 键 词：红外焦平面阵列 非均匀性校正 神经网络校正法 目标退化 伪像 

分 类 号：TN215[电子电信—物理电子学]