基于NSST域非线性扩散滤波修正的人脸偏振热像融合算法  

A Fusion Algorithm of PolarimetricFacial Thermal Images Based on Modified Nonlinear Diffusion Filter in NSST Domain

作　　者：汪方斌 唐晟[1,2] WANG Fangbin;TANG Sheng(School of Mechanical and Electrical Engineering,Anhui Jianzhu University,Hefei 230031,China;Key Laboratory of Construction Machinery Fault Diagnosis and Early Warning,Anhui Jianzhu University,Hefei 230031,China;Key Laboratory of Intelligent Manufacturing of Construction Machinery,Anhui Jianzhu University,Hefei 230031,China)

机构地区：[1]安徽建筑大学机械与电气工程学院,合肥230601 [2]安徽建筑大学建筑机械故障诊断与预警重点实验室,合肥230601 [3]安徽建筑大学工程机械智能制造安徽省重点实验室,合肥230601

出　　处：《长春大学学报》2023年第2期11-20,共10页Journal of Changchun University

基　　金：安徽省教育厅研究项目(YJS20210512);安徽省教育厅高校自然科学重点项目(KJ2020A0487,GXXT-2021-010)。

摘　　要：为使边缘特征与纹理细节更加突出,先利用非负矩阵将Stokes参量图像Q、U分别与I融合;然后利用NSST分解为高频与低频;再采用修正的非线性扩散滤波滤除高频子带噪声;最后以区域能量最大和匹配规则进行高低频系数融合,并通过剪切波逆变换获得最终融合图像。实验表明,基于NSST域非线性扩散滤波修正的融合算法可增强人脸轮廓、边缘和纹理信息,平均梯度、信息熵、标准差、空间频率等评价指标显著提升,视觉效果好。In order to make the edge features and texture be more prominent,firstly,the stokes parameter images Q and U were fused with I through non-negative matrix factorization respectively;secondly,the fused images were decomposed into high-frequency and low-frequency components by NSST;thirdly,the high-frequency sub-band noise was smoothed with the modified nonlinear diffusion filter algorithm;finally,the high and low-frequency subband coefficients were fused with the region energy maximum and energy matching rules,and the final fused image were obtained by inverse NSST.The experimental results showed that this algorithm could enhance the information on the facial contour,edge and texture,the evaluation indexes such as average gradient,information entropy,standard deviation and spatial frequency had been significantly improved,representing good vision effect.

关 键 词：图像融合 偏振热像 扩散滤波 

分 类 号：O436[机械工程—光学工程]