采用特征直方图的红外焦平面阵列盲元检测方法  

Blind-pixel detection algorithm for IRFPA by applying pixel's characteristics histogram analysis

作　　者：张红辉[1,2] 罗海波[1,3,4] 余新荣 丁庆海[6] 

机构地区：[1]中国科学院沈阳自动化研究所,辽宁沈阳110016 [2]中国科学院大学,北京100049 [3]中国科学院光电信息处理重点实验室,辽宁沈阳110016 [4]图像理解与视觉计算重点实验室,辽宁沈阳110016 [5]中航工业江西洪都航空工业集团有限责任公司,江西南昌330024 [6]空军装备研究院总体所,北京100076

出　　处：《红外与激光工程》2014年第6期1807-1811,共5页Infrared and Laser Engineering

基　　金：国家自然科学基金(51005229)

摘　　要：红外探测器受材料和工艺等的限制,普遍存在盲元问题,降低了红外成像系统的图像质量,影响了目标检测系统的检测概率和虚警率,因此,有效检测盲元是红外图像处理领域的一个重要研究方向。在分析了有效像元校正模型的基础上,指出均匀辐照下红外焦平面阵列(IRFPA)有效像元非均匀校正后灰度图像的均值(MEAN)和标准差(STD)具有正态分布特征,对特征直方图采用自适应正交投影分解法进行高斯分解,得到有效像元校正后的MEAN和STD特征分布区间,最终得到盲元的分类准则。对IRFPA进行的验证实验结果表明,该方法获得了令人满意的结果,证实了所提出检测方法的有效性和科学性。Due to the influence of material, fabrication technology and so on, blind-pixels exist on infrared focal plane array detectors. It reduces the infrered image quality and object detection capability, improves the object inveracious rate. Effective blind-pixels detection for infrared image process is an important development issue. Based on the analysis of the model for the effective pixel corrected response, the approach was proposed that, when the corrected response was uniform, the means （MEAN） and the standard deviation （STD） were both normal distributed random variables for all effective pixels. Through adaptive projective decomposition of MEAN and STD histograms, the MEAN and STD distribution range of effective pixel could be effectively estimated, and the detection rule for blind-pixel was got finally. The experimental result by applying real infrared image shows the proposed method for IRFPA blind-pixel detection is scientific and effective.

关 键 词：盲元检测 响应特征 直方图分解 

分 类 号：TN215[电子电信—物理电子学]