摘要
针对红外焦平面成像系统存在列向条纹非均匀性的现象,采用了一种基于自适应PM扩散模型的非均匀校正新算法。首先,综合利用图像梯度信息和局部灰度统计信息,自适应计算PM模型的扩散阈值;然后将每列像素的PM模型估计值作为该列像素的期望值;最后采用最陡下降法迭代计算得到每列像元的校正参数,并对结果进行循环校正以提高校正效果。实验结果表明:该算法可以保护图像边缘信息,与同类算法相比,能够更有效地抑制条纹非均匀性,并且能够防止图像产生鬼影。
In order to correct the stripe nonuniformity for infrared images captured by infraed focal plane array （IRFPA） , a novel stripe nonuniformity correction algorithm based on adaptive PM diffusion models for single inflared image is adopted. Firstly, the adaptive diffusion threshold of PM model is calculated by gradient infor- mation and local gray level statistics of infrared images. Then, the estimate values of each column pixel are treated as expectations, which are in constraint of PM models. Finally, con＇ection parameters in iteration are obtained by method of steepest descent, and the image is corrected repeatedly to improve correction perform- anee. Experimental results indicate that the adopted algorithm can preserve edge information. Compared with other four algorithms, the proposed algorithm has advantage of reducing stripe nonuniformity and removing ghosting artifact.
出处
《中国光学》
EI
CAS
CSCD
2016年第1期106-113,共8页
Chinese Optics
基金
国家自然科学基金项目(No.61203189)
二炮院校青年基金资助项目(No.2014QNJJ023)~~
关键词
单帧红外图像
条纹非均匀校正
自适应扩散模型
最陡下降法
鬼影
single infrared image
stripe nonuniformity correction
adaptive diffusion model
steepest descentmethod
ghosting artifact
作者简介
陈世伟（1979－），男，河北南和人，博士研究生，讲师，2006年于第二炮兵工程大学获得硕士学位，主要从事机器视觉及自动控制方面的研究。E-mail：cshw3876@tom．com
