简易偏振成像系统及应用  

A Simple Polarization Imaging System and Its Application

作　　者：杨敬钰 徐吉 李坤[2] 胡浩丰[3] 郭鑫 Yang Jingyu;Xu Ji;Li Kun;Hu Haofeng;Guo Xin(School of Electrical and Information Engineering,Tianjin University,Tianjin 300072,China;College of Intelligence and Computing,Tianjin University,Tianjin 300350,China;School of Precision Instruments and Opto-Electronics Engineering,Tianjin University,Tianjin 300072,China)

机构地区：[1]天津大学电气自动化与信息工程学院,天津300072 [2]天津大学智能与计算学部,天津300350 [3]天津大学精密仪器与光电子工程学院,天津300072

出　　处：《天津大学学报（自然科学与工程技术版）》2020年第3期237-242,共6页Journal of Tianjin University：Science and Technology

基　　金：深圳孔雀计划资助项目(KQTD20140630115140843);国家自然科学基金资助项目(61571322,61771339)~~

摘　　要：由偏振图像获取的偏振信息可用于图像增强与场景检测,近年来受到科研界与工业界的广泛关注.传统的偏振图像捕捉方法主要是通过旋转彩色相机前摆放的偏振片来采集多个偏振角度的图像或者使用偏振相机捕捉图像.前者只适合采集静态场景且手动换偏振片时容易造成相机镜头的抖动而无法对齐各个偏振角的图像,后者,即特制的偏振相机,由于制作工艺复杂,因此价格高昂.偏振图像采集设备制约了该领域的研究与发展.基于上述原因,搭建了由4个普通彩色相机组成的廉价偏振相机系统,该系统在4个相机前分别前置4个偏振角度的偏振片,即0°、45°、90°、135°,并固定在特制的相机架上,用以保证系统中相机之间的相对位置不变.四相机组成同步电路连接到电脑,利用软件控制同步采集场景偏振图像,减少人工触碰相机引起的误差.为了对齐不同偏振角度的图像获取场景偏振信息,采用多分辨率变分光流方法,利用梯度恒常模型对光照不敏感的特性,计算视图间的密集流场,利用斯托克斯矢量求解场景图像的偏振信息,最后,为了测量提出系统的性能,给出了该偏振系统的一种应用--低曝光图像增强,与传统彩色图像增强相比,在场景极暗处,偏振图像可以区分物体与背景,恢复图像细节,取得了更好的恢复效果.Polarization information obtained from polarization images,which has received extensive attention from scientific research and industrial circles in recent years,can be used for image enhancement and scene detection.The traditional methods to capture a polarization image are mainly to collect images with multiple polarization angles by rotating the polarizer placed in front of the color camera or to use a use polarization camera.The former is only suitable for static scene acquisition and easily causes camera lens movements when the polarizer is changed manually.Alternatively,a special polarization camera is expensive because of its complex manufacturing process,which restricts the research and development in this field.Based on the above reasons,an inexpensive polarization camera system composed of four ordinary color cameras was built.The system has four polarization angles in front of the four cameras,0°,45°,90° and 135°.The system is fixed on a special camera rack to ensure that the relative positions of the cameras in the system remain unchanged.The four cameras are connected to the computer by a synchronous circuit.The software is used to control the synchronous acquisition of the polarization images in the scene to reduce the error caused by the manual touch of the camera.In order to obtain the polarization information of the scene by aligning images with different polarization angles,a multi-resolution variational optical flow method was adopted.The gradient constant model was used to calculate the dense flow field between views.The Stokes vector was used to solve the polarization information of the scene images.Finally,in order to measure the performance of the proposed system,a low exposure application of the polarization system is provided.Compared with traditional color image enhancement,the polarization image can distinguish object from the background,restore image details and achieve better restoration results in extremely dark scenes.

关 键 词：偏振相机 低照度图像 图像恢复 图像增强 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]