-
光的偏振可以用电矢量法、琼斯矩阵法、Poincare球和Stokes矢量法来描述。在偏振成像技术的实际应用中,最为常见的是通过获取Stokes矢量求取需要的偏振特征参数,故重点介绍偏振光的Stokes矢量表示法。
-
Stokes矢量法通过简单的数学表达形式描述光的偏振态。Stokes矢量包含四个分量,各个分量可通过测量相位或特定角度的光强得到。然而,光的频率变化约为1014 Hz,现有探测器的帧频最高为百万赫兹级,无法直接获取光的相位信息。因此,通过采集特定角度(通常选取0°、±45°、90°方向)的偏振子图像并对其强度信息处理得到Stokes矢量,表示为:
$$ S = \left[ {\begin{array}{*{20}{c}} {\begin{array}{*{20}{c}} {\begin{array}{*{20}{c}} {{S_0}} \\ {{S_1}} \end{array}} \\ {{S_2}} \end{array}} \\ {{S_3}} \end{array}} \right] = \left[ {\begin{array}{*{20}{c}} {\begin{array}{*{20}{c}} {\begin{array}{*{20}{c}} {{I_{\text{0}}} + {I_{{\text{90}}}}} \\ {{I_{\text{0}}} - {I_{{\text{90}}}}} \end{array}} \\ {{I_{{\text{45}}}} - {I_{{{ - 45}}}}} \end{array}} \\ {{I_R} - {I_L}} \end{array}} \right] $$ (1) Stokes矢量S=[S0, S1, S2, S3]T可表示任意偏振光的偏振态。其中,S0表示光场总强度,S1表示0°和90°方向线偏振光的光强差,S2表示±45°方向线偏振光的光强差,S3表示左旋与右旋圆偏振光的光强差。偏振度DoP和偏振角β是用来描述偏振光特性的重要参数,用于偏振光学成像,可根据Stokes矢量计算得到,如公式(2)所示:
$$ \begin{gathered} {\rm{ DoP}} = \frac{{\sqrt {S_1^2 + S_2^2{\text{ + }}S_3^2} }}{{{S_0}}}{\text{ }} \\ {\text{ }}\beta = \frac{1}{2}\arctan \left( {\frac{{{{{{S}}}_{\text{2}}}}}{{{{{{S}}}_{\text{1}}}}}} \right) \\ \end{gathered} $$ (2) -
非偏振光入射到目标表面后的传播模型如图1所示,入射光可以分解成垂直和平行于入射平面的分量,菲涅耳方程给出了垂直(平行)于入射平面的线偏振光的反射光振幅与入射光振幅之比[18],基于此,求得非偏振光入射到目标表面后反射光和折射光的偏振度,分别表示为公式(3)、(4),其中,n表示空气中目标表面材料的折射率。
$$ {\rm{DoP}}{_r} = \frac{{\sqrt {{{\sin }^4}\theta {{\cos }^2}\theta \left( {{n^2} - {{\sin }^2}\theta } \right)} }}{{\left[ {{{\sin }^4}\theta + {{\cos }^2}\theta \left( {{n^2} - {{\sin }^2}\theta } \right)} \right]/2}} $$ (3) $$ \begin{gathered} {\rm{DOP}}{_t} = \\ \dfrac{{{{\left( {n -1/n } \right)}^2}{{\sin }^2}\theta }}{{2 + 2{n^2} - {{\left( {n + 1/n} \right)}^2}{{\sin }^2}\theta + 4\cos \theta \sqrt {{n^2} - {{\sin }^2}\theta } }} \\ \end{gathered} $$ (4) -
光照射到各向异性的物质表面时,会产生镜面反射光和漫反射光[50]。根据反射光成分的不同,偏振三维成像可分为基于镜面反射光和漫反射光的偏振三维成像。
物体表面法线方向由天顶角θ(入射角)和入射平面的方位角φ共同决定。图3(a)、3(b)分别给出镜面反射光、漫反射光的偏振度和天顶角的函数关系。可见,基于镜面反射光的偏振三维成像技术存在天顶角不确定的问题,需要进行去模糊处理获取准确的天顶角,将在4.2.1节中详细阐述。
由马吕斯定律知,探测器收集到的光强随偏振器件的旋转而变化。公式(5)给出偏振器件的旋转角度与光强值的变化关系,其关系曲线如图3(c)所示。目标像素无论是以镜面反射光为主还是以漫反射光为主,其法线方位角的实际值与计算值均存在180°的不确定性[51],导致物体表面法线方向不准确,三维面形恢复出现严重畸变。为了得到准确的目标表面法线场,需要对方位角进行去模糊处理。针对镜面反射和漫反射的偏振三维成像技术中的方位角模糊问题,将分别在4.2.2节和4.3节进行介绍。
$$ I = \frac{{{I_{\max }} + {I_{\min }}}}{2} + \frac{{{I_{\max }} - {I_{\min }}}}{2}\cos (2\alpha - 2\varphi ) $$ (5) 式中:Imax、Imin分别表示光强的最大值和最小值;α为偏振片旋转角度。
通过对天顶角和方位角的去模糊处理,可得到唯一的目标表面法线场,如公式(6)所示,选取适当的算法对表面法线场进行积分,能实现目标的三维重建。
$$ n = \left( {\tan \theta \cos \varphi ,\tan \theta \sin \varphi ,1} \right) $$ (6) 实际应用中,对于金属和透明物体,镜面反射特性更为明显,故只需要考虑镜面反射光。而对于非透明电介质目标,表面反射光以漫反射光为主,但也会受到镜面反射光的影响,需要分离镜面反射光,以获得更好的重建效果。
-
2002年,D. Miyazak等人[52]采用旋转目标测量法解决天顶角模糊问题。由图3(a)可知,当天顶角θ=0°或90°时,线偏振度为0,当天顶角为布儒斯特角时,线偏振度为1。如图4(a)~(b)所示,利用布儒斯特曲线将物体表面划分为三个区域,分别为B-E、B-B、B-N,可对天顶角进行分区域消歧。如果物体是封闭光滑的,那么解决特定区域内某一个点的模糊问题,即可完成全区域消歧。同年,D. Miyazak发现当使用红外光照明时,镜面反射光偏振度与天顶角的函数关系是单调的,如图4(c)所示,此时,通过测得光的偏振度可唯一确定天顶角。但是红外光的偏振度明显比可见光小,对于较小的入射角,偏振度很难测量。因此,将可见光和红外光相结合是处理天顶角模糊问题的有效手段[53]。2012年,C. Stolz等[54]提出用多光谱偏振处理方法得到准确的天顶角。图4(d)给出不同波段偏振度与入射角的关系,由于折射率对波长的依赖,布儒斯特角会有一定的改变,根据不同波长光照下偏振度和布儒斯特角间的差异性解决天顶角的模糊问题。实验证明,该方法能够有效重建透明物体的三维形貌。但在实际应用中,需要测量多个波段的强度值,实验装置复杂,具有一定的局限性。2015年,G. Missael等[55]提出利用圆偏振的方法处理天顶角的模糊问题。由图4(e)可知,天顶角与圆偏振度的关系为单调函数,由圆偏振度可唯一确定天顶角,解决天顶角的模糊问题。
图 4 (a)(b) 布儒斯特分割[52];(c) 红外光和可见光下入射角与偏振度的关系[53];(d) 两个不同波长下入射角和偏振度关系图[54];(e) 线偏振度(DoLP)、圆偏振度(DoCP)和天顶角的关系[55]。
Figure 4. (a)(b) Brewster segmentation[52]; (c) The relationship between incident angle and polarization in infrared and visible light[53]; (d) Two degree of polarization curves simulated for two wavelengths[54]; (e) DoLP and DoCP as a function of the zenith angle[55]
-
2006年,O. Morel等[56]提出利用主动照明法消除方位角歧义,采用LED环状光源照明,为物体提供均匀的非偏振光。该光源由四部分组成,可独立控制,分别从四个方向拍摄目标,通过分析各方向的强度图像确定方位角。该方法成像过程较为复杂,且对光源和环境的要求严格。2017年,D. Miyazak等[57]提出利用偏振分析和空间雕刻法恢复目标三维形貌。首先,通过空间雕刻技术粗略估计物体的三维形貌,然后加入偏振信息实现多视角偏振三维探测。该方法充分融合空间雕刻和偏振成像方法的优势,利用奇异值分解(SVD)计算曲面法向量,使最小二乘误差最小化。可用来估计光滑物体的形状,如塑料和陶瓷物体,以及黑色和具有高镜面反射特征的彩色物体。
-
根据4.1节介绍,基于漫反射的偏振三维成像技术不存在天顶角模糊问题,因此只对方位角的消歧方法展开综述。
-
1) 结合光度立体视觉法的偏振三维成像
2007年,G. Atkinson等[58]提出利用光度立体视觉技术对方位角进行消歧。通过比较三个照明角度下拍摄图像光强的大小,实现对方位角的消歧。基于光度立体视觉的偏振三维成像技术虽然能够实现对目标形貌的三维重建,但是该方法对光源的位置要求严格,成像系统较为复杂,不易实现。
2) 结合飞行时间法(TOF)的偏振三维成像
2017年,A. Kadambi等[59]将偏振信息与飞行时间法相结合解决方位角的模糊问题。首先由Kinect(TOF相机)得到的粗糙深度获取表面法线信息Ndepth,然后结合公式(7)、(8)校正由偏振信息得到的表面法线场Npolar。2019年,北京大学杨锦发等[60]利用Astra3D相机(采用红外散斑结构光的方式)获取目标的粗糙深度图,并与偏振信息融合对方位角进行消歧,实现对光滑低纹理目标的三维重建。该方法仅适用于反射成分为漫反射的物体,应用范围有一定的局限性。此外,该方法引入了图像配准的问题,增加了三维重建的复杂度。
$$ \overset{\lower0.5em\hbox{$\smash{\scriptscriptstyle\frown}$}}{A} = \arg {\min _A}\left\| {{N^{{\rm{depth}}}} - A({N^{{\rm{polar}}}})} \right\|_2^2 A \in \left\{ {\begin{array}{*{20}{c}} { - 1}&1 \end{array}} \right\} $$ (7) $$ {N^{{\rm{corr}}}} = \overset{\lower0.5em\hbox{$\smash{\scriptscriptstyle\frown}$}}{A} ({N^{{\rm{polar}}}}) $$ (8) 3) 结合多目立体视觉法的偏振三维成像
2017年,西北工业大学平茜茜等[61]将偏振信息与双目立体视觉相结合,利用双目立体视觉法标定得到相机参数,将偏振得到的图像像素坐标系下的点云数据转化为世界坐标系下的绝对数据,实现了高反光无纹理目标真实深度的测量。2019年,D. Zhu等[62]提出偏振相机和RGB相机的混合探测系统。首先根据两相机的视差生成一个粗糙的深度图,通过计算粗糙深度图的梯度计算引导表面法线,再利用引导表面法线消除由偏振信息获取的表面法线的歧义。2021年,北京大学张瑞华[63]等采用多视角立体几何与偏振信息融合的三维重建算法消除了方位角歧义,并采用泊松优化方法纠正天顶角偏差,实现对低纹理目标形貌的三维重建。2022年,武汉大学田昕等[64]采用拟合数据项描述偏振面与融合结果之间的线性关系,将目标纹理从偏振曲面转移到融合深度中;采用鲁棒低秩矩阵分解约束双目深度和融合深度,有效地考虑了由于像素不匹配导致的缺失项和环境噪声引起的异常值的影响,提高了融合深度的精度,成像效果如图5(a)所示,但是该方法的偏振成像系统较为复杂。
图 5 成像结果。 (a) 基于偏振成像与双目立体视觉融合的三维重建[64];(b) 近红外单目偏振三维成像[66];(c) 基于稀疏线性方程组的线性深度估计[69-70];(d) 基于深度学习的偏振三维重建[72]
Figure 5. Imaging result. (a) 3D reconstruction based on fusion of polarization imaging and binocular stereo vision[64]; (b) Near-infrared monocular polarization 3D imaging[66]; (c) Linear depth estimation based on a sparse system of linear equations[69-70]; (d) Polarization 3D reconstruction based on deep learning[72]
4) 结合结构光投影的偏振三维成像
2017年,浙江大学汪凯巍等[65]采用液晶投影仪(LCD),通过在液晶两端施加不同强度的电压可快速获得具有不同偏振方向的出射光,无需旋转线偏振片进行偏振调制。该方法通过对每个结构光图的快照估计场景中的线偏振度(DoLP),通过DoLP来识别目标,并有选择性地进行重建。该方法有利于高效的三维重建和偏振目标增强。
-
2021年,西安电子科技大学韩平丽等[66]提出一种近红外单目偏振三维成像技术,该方法可直接重建非均匀反射表面的形状。在权重约束中引入参考梯度场,对非均匀反射目标表面法线的模糊进行全局校正。实验证明,该方法可以成功重构出近场和远场反射不均匀的目标形状,如图5(b)所示,并将偏振三维成像的应用扩展到复杂光照条件和较长的探测距离,分辨率为微米级。同年,西北工业大学李磊磊等[67]建立红外偏振辐射模型,该方法不依赖光照条件和目标表面的纹理特征,具有重建精度较高、实时性好和无数据空洞等优点。
-
2012年,A. Mahmoud等[68]提出将阴影恢复法与偏振信息相结合对目标实现三维重建。首先,利用偏振信息得到模糊的方位角,以可能的方位角为元素构成集合R1={φ,φ+π},然后,根据阴影信息得到的方位角构成的集合为R1={φ1,φ2}。
$$ \sum\limits_{m = {\text{1}}}^{\text{2}} {d(r,{R_m}) = } \sum\limits_{m = 1}^{\text{2}} {{{\min }_{j = 1, \ldots ,{n_m}}}\left| {r - {r_{mj}}} \right|} $$ (9) 式中:Rm(m=1,2)表示这两组中的任何一组,将集合中的元素与任意给定的r值进行比较,使得公式(9)值最小,则该元素为方位角的值。该方法假设目标表面是漫反射表面,对镜面反射像素并未处理,应用具有一定的局限性。2019年,W. Smith等[69-70]等提出通过求解大型稀疏线性方程组从单帧偏振图像中恢复表面高度。不同于其他利用偏振信息恢复表面高度的方法,该方法不需要单独进行方位角去模糊处理,因为在求解深度的线性方程组时,方位角模糊以全局最优解的方式得以解决。该方法在已知光源方向和目标表面均匀反射的情况下,首先对表面梯度进行平滑中心差分近似,然后将偏振约束和阴影约束表示为与未知深度相关的大型稀疏线性方程组的形式,最后利用线性最小二乘法对高度进行优化。W. Smith将该方法其扩展到一个未校准的室外场景,对不同材料的物体形貌均能实现三维重建,如图5(c)所示。2022年,该团队利用独立成分分析的算法将镜面反射和漫反射进行分离,然后利用朗伯体反射模型将漫反射光的强度数据转换为高度数据,再根据高度信息得到表面法线信息,最后利用公式(7)和公式(8)进行校正[71]。实验表明该方法可以达到微米级的深度分辨率。
-
2020年,Y. Ba等[72]提出深度学习结合偏振信息的方法实现目标三维重建,如图5(d)所示。该模型将0°、45°、90°、135°的偏振图像和模糊法线作为输入,通过神经网络学习,最终输出准确的表面法线。2022年,西安电子科技大学韩平丽等[73]采用基于卷积神经网络的3DMM (3D Morphable Model)模型获取每一像素的模糊表面法线,对由偏振信息得到的表面法线进行约束,从而实现了在自然光照明的环境中实现了人脸的三维重建。
目前,偏振三维成像技术已经能够实现对单一静态目标的三维重建。然而,基于镜面反射光的偏振三维成像技术中天顶角的消歧过程繁琐,无法通过一次探测确定唯一的天顶角;基于漫反射光的偏振三维成像技术存在漫反射光分量少,不易探测和镜面反射光干扰等问题,针对其方位角的模糊问题,通常需要结合其他三维感知技术获取先验信息,实现对方位角的约束,严重制约了基于漫反射光偏振三维成像技术的广泛应用。
Research, application and progress of optical polarization imaging technology
-
摘要: 偏振成像技术作为一种新型的光学成像技术,可以实现抑制背景噪声、提高探测距离、获取目标细节特征和识别伪装目标等功能。由于成像空间维度的不同,偏振二维成像和偏振三维成像在不同领域中具有良好的应用前景。文中从偏振光的表示与传播方式入手,先后对偏振成像系统、偏振二维成像技术、偏振三维成像技术和基于超表面偏振器件的偏振探测及成像的研究展开综述。首先,根据偏振成像系统结构的不同,偏振成像系统可分为分时型、分振幅型、分孔径型和分焦平面型四种,并对以上偏振成像系统分别进行详细介绍和比较分析。其次,阐述了基于图像增强技术的偏振二维成像。图像增强技术分为偏振差分算法和图像融合两种。对于偏振三维成像,根据所处理反射光成分的不同,分为基于镜面反射光和漫反射光的偏振三维成像。综述了三维形貌重建过程中天顶角和方位角多值性问题的解决办法。为了高效准确地获取偏振信息,基于超表面结构的偏振器件成为当前研究的热点。进一步介绍了基于超表面偏振器件的偏振探测及成像技术。最后,总结全文并对偏振成像技术的发展前景进行展望。Abstract:
Significance Traditionally, light intensity was utilized in optical imaging, resulting in multi-dimensional physical quantities such as spectrum, polarization and phase, and the light field information are lost, which lead to the poor performance or even failure of the traditional method in harsh conditions. However, polarization imaging technology utilizes the polarization property of light, which is insensitive to background illumination, ambient temperature and contrast. Meanwhile, polarization characteristic of light can be reserved more probably in low than that of the light intensity, and so it is more applicable to achieve effective detection of targets in special environments. Based on the unique advantages of polarization imaging, the technology is widely used in the fields of communication, imaging and detection. Progress Firstly, four types of traditional polarization imaging systems of time-sharing (TS), division-of-amplitude (DOA), division-of-aperture (DoAp) and division-of-focal-plane (DoFP) are introduced. Except for the TS polarizaition imaging method, the other three methods all performed well in real-time imaging. The TS polarization imaging system is simple in structure and is commonly used in polarization differential imaging and 3D imaging. The DOA polarization imaging system is relatively complex and difficult to calibrate, resulting in its poor practicality. Structure of DoAp polarization imaging system is relatively compact, but the image alignment is relatively complicated. The DoFP polarization imaging system became a focus in recent researches, owing to its advantages of low energy loss, compact structure and fast imaging. For this technology, low extinction ratio of the micro-polarization array produced during the fabrication process was significantly improved with the enormous progress in processing technology and this approach is most likely to be predominant in future polarization imaging.Based on the traditional polarization imaging system, polarization 2D/3D imaging technology has been studied and made great progress. Based on polarization difference and image fusion, the polarization 2D imaging technique that has achieved good imaging results in underwater and haze environments is illustrated in detail. 2D imaging through strong scattering media and separation of high and low polarization targets are still challenging at present. For polarization 3D imaging technology, this paper provides a detailed description of the methods to solve the azimuth and zenith angle multivalence problems in the imaging process. Although high-precision 3D reconstruction of a single object in the natural environment is currently possible, the relative height of the target rather than the absolute height is recovered (Fig.5). In addition, with the existing polarization 3D imaging technology, it is unable to achieve 3D shape recovery for discontinuous and dynamic targets, and further research is still indispensable to solve these problems.With the development of micro-nano processing and integration technology, smaller and more integrated metasurface structures have been studied and applied to polarization detection. At present, full polarization detection, and polarization imaging has been realized by using polarization devices based metasurface (Fig.6). Conclusions and Prospects Polarization imaging technology is elaborated in two aspects of polarization detection and imaging. For polarization detection, four traditional polarization imaging systems were introduced respectively. Therein, the DoFP polarization imaging system has drawn more attention due to its unique advantages of fast imaging speed and good integration. Depending on the spatial dimension, polarization 2D/3D imaging based on the traditional polarization imaging system have good prospects for detection and imaging in different fields. In order to achieve polarization detection and imaging more efficiently and conveniently, polarization devices based metasurface are fabricated and applied. Scientists at home and abroad are dedicated to continuously optimize the imaging process from five aspects of generation, transmission, modulation, acquisition and processing of polarized light, and various advanced processes and methods are effectively combined to achieve on-line polarization imaging with high stability. -
Key words:
- optical imaging /
- 2D polarization imaging /
- 3D polarization imaging /
- metasurface
-
图 4 (a)(b) 布儒斯特分割[52];(c) 红外光和可见光下入射角与偏振度的关系[53];(d) 两个不同波长下入射角和偏振度关系图[54];(e) 线偏振度(DoLP)、圆偏振度(DoCP)和天顶角的关系[55]。
Figure 4. (a)(b) Brewster segmentation[52]; (c) The relationship between incident angle and polarization in infrared and visible light[53]; (d) Two degree of polarization curves simulated for two wavelengths[54]; (e) DoLP and DoCP as a function of the zenith angle[55]
图 5 成像结果。 (a) 基于偏振成像与双目立体视觉融合的三维重建[64];(b) 近红外单目偏振三维成像[66];(c) 基于稀疏线性方程组的线性深度估计[69-70];(d) 基于深度学习的偏振三维重建[72]
Figure 5. Imaging result. (a) 3D reconstruction based on fusion of polarization imaging and binocular stereo vision[64]; (b) Near-infrared monocular polarization 3D imaging[66]; (c) Linear depth estimation based on a sparse system of linear equations[69-70]; (d) Polarization 3D reconstruction based on deep learning[72]
-
[1] Short N J, Yuffa A J, Videen G W, et al. Effects of surface materials on polarimetric-thermal measurements: applications to face recognition [J]. Applied Optics, 2016, 55(19): 5226-5233. doi: 10.1364/AO.55.005226 [2] Gurton K P, Yuffa A J, Videen G W. Enhanced facial recognition for thermal imagery using polarimetric imaging [J]. Optics Letters, 2014, 39(13): 3857-3859. doi: 10.1364/OL.39.003857 [3] Wang Y, Yang C, Wang Y, et al. Gigabit polarization division multiplexing in visible light communication [J]. Optics Letters, 2014, 39(7): 1823-1826. doi: 10.1364/OL.39.001823 [4] Gaiarin S, Perego A, Silva E, et al. Dual-polarization nonlinear Fourier transform-based optical communication system [J]. Optica, 2018, 5(3): 263-270. doi: 10.1364/OPTICA.5.000263 [5] He C, He H, Chang J, et al. Polarisation optics for biomedical and clinical applications: a review [J]. Light: Science & Applications, 2021, 10: 194. doi: 10.1038/s41377-021-00639-x [6] Liu Y, He H, Wu J. Differentiation of human GBM from non-gbm brain tissue with polarization imaging technique [J]. Frontiers in Oncology, 2022, 12: 863682. doi: 10.3389/fonc.2022.863682 [7] Kim M, Keller D, Bustamante C. Differential polarization imaging. I. Theory [J]. Biophysical Journal, 1987, 52(6): 911-927. doi: 10.1016/S0006-3495(87)83285-X [8] Rowe M P, Pugh E N, Tyo J S, et al. Polarization-difference imaging a biologically inspired technique for observation through scattering media [J]. Optics Letters, 1995, 20(6): 608-610. doi: 10.1364/OL.20.000608 [9] 陈伟力, 王霞, 金伟其, 等. 采用中波红外偏振成像的目标探测实验[J]. 红外与激光工程, 2011, 40(1): 7-11. doi: 10.3969/j.issn.1007-2276.2011.01.002 Chen Weili, Wang Xia, Jin Weiqi, et al. Experiment of target detection based on medium infrared polarization imaging [J]. Infrared and Laser Engineering, 2011, 40(1): 7-11. (in Chinese) doi: 10.3969/j.issn.1007-2276.2011.01.002 [10] Yang F, Wei H. Fusion of infrared polarization and intensity images using support value transform and fuzzy combination rules [J]. Infrared Physics & Technology, 2013, 60: 235-243. doi: 10.1016/j.infrared.2013.05.008 [11] Hu H, Zhao L, Li X, et al. Polarimetric image recovery in turbid media employing circularly polarized light [J]. Optics Express, 2018, 26(19): 25047-25059. doi: 10.1364/OE.26.025047 [12] Liu F, Wei Y, Han P, et al. Polarization-based exploration for clear underwater vision in natural illumination [J]. Optics Express, 2019, 27(3): 3629-3641. doi: 10.1364/OE.27.003629 [13] Liu F, Cao L, Shao X, et al. Polarimetric dehazing utilizing spatial frequency segregation of images [J]. Applied Optics, 2015, 54(27): 8116-8122. doi: 10.1364/AO.54.008116 [14] Qu Y, Zou Z. Non-sky polarization-based dehazing algorithm for non-specular objects using polarization difference and global scene feature [J]. Optics Express, 2017, 25(21): 25004-25022. doi: 10.1364/OE.25.025004 [15] Frankot R T, Chellappa R. A method for enforcing integrability in shape from shading algorithms [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1988, 10(4): 439-451. doi: 10.1109/34.3909 [16] 赵群, 李志宏, 杨进华. 基于松弛迭代法实现物体三维结构的重建[J]. 长春理工大学学报, 2008, 31(2): 7-10. doi: 10.3969/j.issn.1672-9870.2008.02.003 Zhao Q, Li Z H, Yang J H. 3D surface reconstruction base on relaxation method [J]. Journal of Changchun University of Science and Technology (Natural Science Edition), 2008, 31(2): 7-10. (in Chinese) doi: 10.3969/j.issn.1672-9870.2008.02.003 [17] Kovesi P. Shapelets correlated with surface normals produce surfaces[C]//10th IEEE International Conference on Computer Vision (ICCV 2005), 2005: 994-1001. [18] Atkinson G A, Hancock E R. Atkinson E R H. Recovery of surface orientation from diffuse polarization [J]. IEEE Transactions on Image Processing, 2006, 15(6): 1653-1664. doi: 10.1109/TIP.2006.871114 [19] Peterson J Q, Jensena G L, Kristib J A, et al. Polarimetric imaging using a continuously spinning polarizer element[C]// Polarization Analysis, Measurement, and Remote Sensing III, 2000, 4133: 292-300. [20] Bigué L, Cheney N. High-speed portable polarimeter using a ferroelectric liquid crystal modulator[C]//Conference on Polarization Science and Remote Sensing III, 2007, 6682: 668205. [21] Gendre L, Foulonneau A, Bigué L. Stokes imaging polarimetry using a single ferroelectric liquid crystal modulator[C]//Proceedings of SPIE, 2010, 7672: 76720B. [22] Azzam R M A. Division-of-amplitude Photopolarimeter (DOAP) for the simultaneous measurement of all four stokes parameters of light [J]. Optica Acta:International Journal of Optics, 1982, 29(5): 685-689. doi: 10.1080/713820903 [23] Phenis A M, Virgen M, Leon E E. Achromatic instantaneous Stokes imaging polarimeter[C]//Novel Optical Systems Design and Optimization VIII, 2005, 5875: 587502. [24] Mudge J, Virgen M, Dean P. Near-infrared simultaneous Stokes imaging polarimeter[C]//Conference on Polarization Science and Remote Sensing IV, 2009, 7461: 74610L. [25] Mudge J, Virgen M. Near-infrared simultaneous Stokes imaging polarimeter: integration, field acquisitions, and instrument error estimation[C]//Conference on Polarization Science and Remote Sensing V, 2011, 8160: 81600B. [26] Wang Y, Hu X, Lian J, et al. Geometric calibration algorithm of polarization camera using planar patterns [J]. Journal of Shanghai Jiaotong University (Science), 2017, 22(1): 55-59. doi: 10.1007/s12204-017-1799-3 [27] Pezzaniti J L, Chenault D B. A division of aperture MWIR imaging polarimeter[C]//Polarization Science and Remote Sensing II, 2005, 5888: 58880V. [28] Liu X, Zhai S, Li J, et al. Design of cooled medium-wave infrared polarization imaging optical system [J]. Infrared and Laser Engineering, 2021, 50(2): 20200208. (in Chinese) doi: 10.3788/irla20200208 [29] Luo H, Zhang J, Gai X, et al. Development status and prospects of polarization imaging technology (Invited) [J]. Infrared and Laser Engineering, 2022, 51(1): 20210987. (in Chinese) doi: 10.3788/IRLA20210987 [30] 罗海波, 刘燕德, 兰乐佳, 等. 分焦平面偏振成像关键技术[J]. 华东交通大学学报, 2017, 34(1): 8-13. Luo Haibo, Liu Yande, Lan Lejia, et al. Key technologies of polarization lmaging for division of focalplane polarimeters [J]. Journal of East China Jiaotong University, 2017, 34(1): 8-13. (in Chinese) [31] Nordin G P, Meier J T, Deguzman P C, et al. Micropolarizer array for infrared imaging polarimetry [J]. Journal of the Optical Society of America A-Optics Image Science and Vision, 1999, 16(5): 1168-1174. doi: 10.1364/JOSAA.16.001168 [32] Perkins R, Gruev V. Signal-to-noise analysis of Stokes parameters in division of focal plane polarimeters [J]. Optics Express, 2010, 18(25): 25815-25824. doi: 10.1364/OE.18.025815 [33] Zhao X, Bermak A, Boussaid F, et al. Liquid-crystal micropolarimeter array for full Stokes polarization imaging in visible spectrum [J]. Optics Express, 2010, 18(17): 17776-17787. doi: 10.1364/OE.18.017776 [34] Rebhan D, Rosenberger M, Notni G. Principle investigations on polarization image sensors[C]//Photonics and Education in Measurement Science 2019, 11144: 111440A. [35] Tyo J S, Rowe M P, Pugh EN, et al. Target detection in optically scattering media by polarization-difference imaging [J]. Applied Optics, 1996, 35(11): 1855-1870. doi: 10.1364/AO.35.001855 [36] Treibitz T, Schechner Y Y. Active polarization descattering [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2009, 31(3): 385-399. doi: 10.1109/TPAMI.2008.85 [37] Huang B, Liu T, Hu H, et al. Underwater image recovery considering polarization effects of objects [J]. Optics Express, 2016, 24(9): 9826-9838. doi: 10.1364/OE.24.009826 [38] Wang J, Wan M, Gu G, et al. Periodic integration-based polarization differential imaging for underwater image restoration [J]. Optics and Lasers in Engineering, 2022, 149: 106785. doi: 10.1016/j.optlaseng.2021.106785 [39] Schechner YY, Narasimhan S G, Nayar S K. Instant dehazing of images using polarization[C]//Conference on Computer Vision and Pattern Recognition, 2001: 325-332. [40] Schechner Y Y, Narasimhan S G, Nayar S K. Polarization-based vision through haze [J]. Applied Optics, 2003, 42(3): 511-525. doi: 10.1364/AO.42.000511 [41] Fang S, Xia X, Huo X, et al. Image dehazing using polarization effects of objects and airlight [J]. Optics Express, 2014, 22(16): 19523-19537. doi: 10.1364/OE.22.019523 [42] Cao L, Shao X, Liu F, et al. Dehazing method through polarimetric imaging and multi-scale analysis[C]//Conference on Satellite Data Compression, Communications, and Processing XI, 2015, 9501: 950111. [43] Lavigne D A, Breton M. A new fusion algorithm for shadow penetration using visible and midwave infrared polarimetric images[C]//2010 13th International Conference on Information Fusion, 2010:1-7. [44] Zhou P, Wang F, Zhang H, et al. Camouflaged target detection based on visible and near infrared polarimetric imagery fusion[C]//International Symposium on Photoelectronic Detection and Imaging 2011: Advances in Imaging Detectors and Applications, 2011, 8194: 81940Y. [45] Liu Z, Zeng H, Wang H, et al. Visible polarization image fusion with non-subsampled Shearlets[C]//International Conference on Frontiers in Optical Imaging Technology and Applications, 2015, 9795: 97951S [46] Liang J, Zhang W, Ren L, et al. Polarimetric dehazing method for visibility improvement based on visible and infrared image fusion [J]. Applied Optics, 2016, 55(29): 8221-8226. doi: 10.1364/AO.55.008221 [47] Yan X, Qin H, Li J, et al. Infrared and visible image fusion using multiscale directional nonlocal means filter [J]. Applied Optics, 2015, 54(13): 4299-4308. doi: 10.1364/AO.54.004299 [48] Zhang J, Shao J, Chen J, et al. Polarization image fusion with self-learned fusion strategy [J]. Pattern Recognition, 2021, 118: 108045. doi: 10.1016/j.patcog.2021.108045 [49] Zhang J, Zhang X, Li T, et al. Visible light polarization image desmogging via cycle convolutional neural network [J]. Multimedia Systems, 2021, 28(1): 45-55. doi: 10.1007/s00530-021-00802-9 [50] Shen H, Zheng Z. Real-time highlight removal using intensity ratio [J]. Applied Optics, 2013, 52(19): 4483-4493. doi: 10.1364/AO.52.004483 [51] Wolff L B, Boult T E. Constraining object features using a polarization reflectance model [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1991, 13(7): 635-657. doi: 10.1109/34.85655 [52] Miyazak D, Kagesawa M, Ikeuchi K. Determining shapes of transparent objects from two polarization images[C]//Proceedings of the IAPR Conference on Machine Vision Applications (IAPR MVA 2002), 2002: 26-31. [53] Miyazaki D, Saito M, Sato Y, et al. Determining surface orientations of transparent objects based on polarization degrees in visible and infrared wavelengths [J]. Journal of the Optical Society of America A-Optics Image Science and Vision, 2002, 19(4): 687-694. doi: 10.1364/JOSAA.19.000687 [54] Stolz C, Ferraton M, Meriaudeau F. Shape from polarization a method for solving zenithal angle ambiguity [J]. Optics Letters, 2012, 37(20): 4218-4220. doi: 10.1364/OL.37.004218 [55] Garcia N M, Erausquin I d, Edmiston C, et al. Surface normal reconstruction using circularly polarized light [J]. Optics Express, 2015, 23(11): 14391-14406. doi: 10.1364/OE.23.014391 [56] Morel O, Stolz C, Meriaudeau F, et al. Active lighting applied to three-dimensional reconstruction of specular metallic surfaces by polarization imaging [J]. Applied Optics, 2006, 45(17): 4062-4068. doi: 10.1364/AO.45.004062 [57] Miyazaki D, Shigetomi T, Baba M, et al. Surface normal estimation of black specular objects from multiview polarization images [J]. Optical Engineering, 2017, 56(4): 1-17. doi: 10.1117/1.Oe.56.4.041303 [58] Gary A. Atkinson E R H. Surface reconstruction using polarization and photometric stereo[C]//12th International Conference on Computer Analysis of Images and Patterns, 2007, 4673(58): 466-473. [59] Kadambi A, Taamazyan V, Shi B, et al. Depth sensing using geometrically constrained polarization normals[C]//IEEE International Conference on Computer Vision (ICCV), 2017. 125(1-3): 34-51. [60] 杨锦发, 晏磊, 赵红颖, 等. 融合粗糙深度信息的低纹理物体偏振三维重建[J]. 红外与毫米波学报, 2019, 38(6): 819-827. doi: 10.11972/j.issn.1001-9014.2019.06.020 Yang J, Yan L, Zhao H, et al. Shape from polarization of low-texture objects with rough depth information [J]. J Infrared Millim Waves, 2019, 38(6): 819-827. (in Chinese) doi: 10.11972/j.issn.1001-9014.2019.06.020 [61] 平茜茜, 刘勇, 董欣明, 等. 基于偏振双目视觉的无纹理高反光目标三维重构[J]. 红外与毫米波学报, 2017, 36(4): 432-438. doi: 10.11972/j.issn.1001-9014.2017.04.009 Ping X, Liu Y, Dong X, et al. 3-D reconstruction of textureless and high-reflective target by polarization and binocular stereo vision [J]. J Infrared Millim Waves, 2017, 36(4): 432-438. (in Chinese) doi: 10.11972/j.issn.1001-9014.2017.04.009 [62] Zhu D, Smith W A P. Depth from a polarisation + RGB stereo pair[C]//IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2019: 7569-7578. [63] 张瑞华, 施柏鑫, 杨锦发, 等. 基于视差角和天顶角优化的偏振多视角三维重建[J]. 红外与毫米波学报, 2021, 40(1): 133-142. doi: 10.11972/j.issn.1001-9014.2021.01.018 Zhang R, Shi B, Yang J, et al. Polarimetric multi-view 3D reconstruction based on parallax angle and zenith angle optimization [J]. J Infrared Millim Waves, 2021, 40(1): 133-142. (in Chinese) doi: 10.11972/j.issn.1001-9014.2021.01.018 [64] Tian X, Liu R, Wang Z, et al. High quality 3D reconstruction based on fusion of polarization imaging and binocular stereo vision [J]. Information Fusion, 2022, 77: 19-28. doi: 10.1016/j.inffus.2021.07.002 [65] Huang X, Bai J, Wang K, et al. Target enhanced 3D reconstruction based on polarization-coded structured light [J]. Optics Express, 2017, 25(2): 1173-1184. doi: 10.1364/OE.25.001173 [66] Li X, Liu F, Shao X. Near-infrared monocular 3D computational polarization imaging of surfaces exhibiting nonuniform reflectance [J]. Optics Express, 2021, 29(10): 15616-15630. doi: 10.1364/OE.423790 [67] 李磊磊, 黄海霞, 郭阳, 等. 基于红外辐射偏振成像的目标三维重建方法[J]. 红外与毫米波学报, 2021, 40(3): 413-419. doi: 10.11972/j.issn.1001-9014.2021.03.018 Leilei L, Haixia H, Yang G, et al. 3D reconstruction method of target based on infrared radiation polarization imaging [J]. J Infrared Millim Waves, 2021, 40(3): 413-419. (in Chinese) doi: 10.11972/j.issn.1001-9014.2021.03.018 [68] Mahmoud A H, El-Melegy M T, Farag A A. Direct method for shape recovery from polarization and shading[C]//19th IEEE International Conference on Image Processing (ICIP), 2012: 1769-1772. [69] Smith WAP, Ramamoorthi R, Tozza S. Linear depth estimation from an uncalibrated, monocular polarisation image[C]//14th European Conference on Computer Vision (ECCV), 2016, 9912: 109-125. [70] Smith W A P, Ramamoorthi R, Tozza S. Height-from-polarisation with unknown lighting or albedo [J]. IEEE Trans Pattern Anal Mach Intell, 2019, 41(12): 2875-2888. doi: 10.1109/TPAMI.2018.2868065 [71] Han P, Cai Y, Liu F, et al. Computational polarization 3D: New solution for monocular shape recovery in natural conditions [J]. Optics and Lasers in Engineering, 2022, 151: 106925. doi: 10.1016/j.optlaseng.2021.106925 [72] Ba Y, Gilbert A, Wang F, et al. Deep shape from polarization[C]//ECCV 2020: Computer Vision – ECCV 2020, 2020, 12369: 554-571. [73] Han P, Li X, Liu F, et al. Accurate passive 3D polarization face reconstruction under complex conditions assisted with deep learning [J]. Photonics, 2022, 9(12): 1-12. doi: 10.3390/photonics9120924 [74] Ding F, Chen Y, Bozhevolnyi S I. Metasurface-based polarimeters [J]. Applied Sciences, 2018, 8(4): 594. doi: 10.3390/app8040594 [75] 倪一博, 闻顺, 沈子程, 等. 基于超构表面的多维光场感知[J]. 中国激光, 2021, 48(19):1-28. doi: 10.3788/CJL202148.1918003 Ni Y B, Wen S, Shen Z C, et al. Multidimensional light field sensing based on metasurfaces [J]. Chinese Journal of Lasers, 2021, 48(19): 1918003. (in Chinese) doi: 10.3788/CJL202148.1918003 [76] Wu P C, Chen J W, Yin C W, et al. Visible metasurfaces for on-chip polarimetry [J]. ACS Photonics, 2018, 5(7): 2568-2573. doi: 10.1021/acsphotonics.7b01527 [77] Basiri A, Chen X, Bai J, et al. Nature-inspired chiral metasurfaces for circular polarization detection and full-Stokes polarimetric measurements [J]. Light: Science & Applications, 2019, 8: 78. doi: 10.1038/s41377-019-0184-48 [78] Li L, Wang J, Kang L, et al. Monolithic full-stokes near-infrared polarimetry with chiral plasmonic metasurface integrated graphene-silicon photodetector [J]. ACS Nano, 2020, 14(12): 16634-16642. doi: 10.1021/acsnano.0c00724 [79] Zhang C, Hu J, Dong Y, et al. High efficiency all-dielectric pixelated metasurface for near-infrared full-Stokes polarization detection [J]. Photonics Research, 2021, 9(4): 583-589. doi: 10.1364/PRJ.415342 [80] Sun X, Ma R, Pu X, et al. High-efficiency polarization multiplexing metalenses [J]. Nanomaterials (Basel), 2022, 12(9): 1-9. doi: 10.3390/nano12091500 [81] Arbabi E, Kamali S M, Arbabi A, et al. Full-stokes imaging polarimetry using dielectric metasurfaces [J]. ACS Photonics, 2018, 5(8): 3132-3140. doi: 10.1021/acsphotonics.8b00362 [82] Rubin N A, D'Aversa G, Chevalier P, et al. Matrix fourier optics enables a compact full-stokes polarization camera [J]. Science, 2019, 365(6448): 1-8. doi: 10.1126/science.aax1839 [83] Rubin N A, Chevalier P, Juhl M, et al. Imaging polarimetry through metasurface polarization gratings [J]. Optics Express, 2022, 30(6): 9389-9412. doi: 10.1364/OE.450941 [84] Cheng B, Zou Y, Shao H, et al. Full-Stokes imaging polarimetry based on a metallic metasurface [J]. Optics Express, 2020, 28(19): 27324-27336. doi: 10.1364/OE.400622