基于结构特征先验与多尺度反锐化掩模机制的红外图像增强算法  

Infrared image enhancement algorithm based on structural feature priori and multi-scale non-sharpening masking mechanism

作　　者：姜维[1] 王学春[1] 杨勇[1] Jiang Wei;Wang Xuechun;Yang Yong(School of Infonnation Engineering,Huanghe Science and Technology College,Zhumadian 450063,China)

机构地区：[1]黄河科技学院信息工程学院,驻马店450063

出　　处：《电子测量与仪器学报》2019年第4期1-9,共9页Journal of Electronic Measurement and Instrumentation

基　　金：国家自然科学基金(61502432);河南省科技攻关项目(152102210001);河南省人力资源与社会保障厅博士后项目(2014022);郑州市"智能图像处理与识别"重点实验室项目(郑[2013]2号)资助

摘　　要：为了解决当前红外图像增强算法难以较好地兼顾边缘增强与噪声抑制,导致增强结果易丢失细节与视觉不自然的问题,提出了基于结构特征先验与多尺度反锐化掩模机制的红外图像增强算法。首先,利用8个边缘核与8个角点核,对Prewitt梯度算子进行拓展,从多个方向来提取红外图像的结构特征映射;随后,利用结构特征映射来计算红外目标的约束控制函数,并基于Gibbs先验模型,构建结构特征先验,获取红外图像对应的最优估计;基于红外图像的多尺度特征,引入贝叶斯函数,并联合最优估计,对红外图像完成有序平滑处理;最后,利用多尺度特征映射来改进传统的反锐化掩模算法,对平滑后的红外图像完成增强。实验结果显示,与当前红外图像增强方案相比,所提算法具有更高的增强质量与噪声抑制能力,可以更好地保持图像细节,其模糊线性指数与熵值分别为0. 21、7. 35。In order to solve the defects of missing detail and vision artificiality induced by difficult to synchronize edge enhancement and noise suppression in current infrared image enhancement algorithm,an infrared image enhancement algorithm based on structural feature prior and multi-scale non-sharpening masking mechanism is proposed in this paper. Firstly,the Prewitt gradient operator was extended by using 8 edge kernels and 8 corner kernels for extracting structural feature mapping of infrared images from multiple directions.Subsequently,the constraint controlling function of infrared target was calculated by structural feature map,and the structure feature prior was constructed based on Gibbs prior model to obtain the optimal estimation of infrared images. The multi-scale feature maps of infrared image were extracted,and the infrared image was smoothed to effectively suppress background noise by introducing Bayesian function and combining structure feature prior. Finally,multi-scale feature mapping is used to improve the traditional non-sharpening masking algorithm to enhance the infrared image. Experimental results show that this algorithm has higher enhancement quality and noise suppression ability,as well as better maintaining image details which linear index of fuzziness and entropy values was 0. 21,7. 35 compared with current infrared image enhancement schemes.

关 键 词：红外图像增强 结构特征先验 特征映射 约束控制函数 多尺度特征 贝叶斯函数 反锐化掩模 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] TNO[自动化与计算机技术—计算机科学与技术]