1.School of Physics and Optoelectronic Engineering, Xidian Univeristy, Xi’an 710071, China 2.Xi’an Key Laboratory of Computational Imaging, Xidian University, Xi’an 710071, China 3.Interdisciplinary Research Center on Advanced Optics and Perception, Xidian University, Xi’an 710071, China
Fund Project:Project supported by the National Natural Science Foundation of China (Grant Nos. 61975254, 62005203, 62075175).
Received Date:13 April 2021
Accepted Date:02 June 2021
Available Online:15 August 2021
Published Online:20 November 2021
Abstract:Owing to the inhomogeneity of the refractive index inside the sample (e.g. biological tissue) or on the surface of the sample(e.g. ground glass), light will be strongly scattered when it propagates through the sample. Therefore, we can hardly obtain the information about the objects behind the scattering medium, except for only a complex speckle pattern. To date, many approaches to realize focusing and imaging through scattering medium have been put forward. The traditional method mainly utilizes ballistic photons for imaging through scattering medium. Since the ballistic light is attenuated exponentially with the increase of depth of propagation in the scattering medium, the reconstruction from the speckle formed by scattered light is more conducive to practicability such as deep biomedical imaging. Typically, the wavefront shaping, optical transmission matrix and speckle correlation techniques which can successfully recover hidden object from the speckle, are valuable in biomedical imaging field. However, both optical transmission matrix and wavefront shaping rely on the coherence of light waves. The physical model of speckle correlation imaging is limited by the similarity of the point spread function of the imaging system. Thus, it is restrictive to achieve imaging through random scattering medium with broadband light illumination by using the current techniques.In this paper, we present a broadband scattering imaging method based on common-mode rejection of polarization characteristic. In order to solve the problem that current scattering imaging methods are limited by the spectral width of the light source illumination, the polarization characteristic of the speckle field is explored in depth. We qualitatively analyze the difference in polarization information between the hidden object and the background noise in the speckle field. Notably, owing to the differences among autocorrelation functions of the speckle field intensity with different rotate angles of polarization, we can obtain two images where the object information contained in the speckle field and the background noise are dominant. Specifically, two speckle patterns are selected according to the maximum value and minimum value of the peak-to-correlation energy of the different speckles’ intensity autocorrelation. Afterwards, the serious background noise caused by the broadband light illumination is significantly suppressed by using polarization speckle difference imaging, and then the hidden object is reconstructed, with basic phase retrieval algorithm combined.Comparison with conventional speckle correlation imaging technique, the value of peak signal-to-noise ratio and structural similarity index of reconstructions through using the proposed method are improved significantly, and the fitting curves are stabilized. Emphatically, the background noise item is physically handled by developing a novel physical imaging model. Furthermore, the proposed method is highly efficient and universal to recover different types of the hidden objects with better quality under broadband light illumination. Therefore, the proposed method has more potential applications in scattering imaging and biomedical imaging. Keywords:broadband scattering imaging/ polarization/ difference imaging
全文HTML
--> --> -->
2.宽谱散射成像模型如图1所示为基于宽谱光源照明的透过散射介质成像系统, 探测器获得的散斑场强度图像I为[12] 图 1 基于宽谱光源照明的透过散射介质成像原理示意图 Figure1. Schematic of imaging through scattering medium with broadband light illumination.
$\begin{split} &I = \int_{ - \infty }^\infty {O\left( r \right)} \cdot S\left( r \right){\rm{d}}r \\=\;& \int_{ - \infty }^\infty {A\left( r \right) \cdot {{\rm{e}}^{{\rm{i}}\varphi \left( r \right)}}} \cdot S\left( r \right){\rm{d}}r = O * S \rm{, } \end{split}$
式中, $O\left( r \right)$为成像目标, $S\left( r \right)$为系统的PSF, r表示空间位置, $A\left( r \right)$和${{\rm{e}}^{{\rm{i}}\varphi \left( r \right)}}$分别为$O\left( r \right)$的振幅与相位, $ * $表示卷积运算. $O\left( r \right)$与$S\left( r \right)$分别简记为O和S. 根据散射介质具有光学记忆效应的物理特性[13-14], 当目标尺寸在光学记忆效应范围内, 成像系统PSF具有空间位移不变性, 目标信息经过散射介质后将形成近乎相同的散斑光场, 则目标的自相关与散斑光场的自相关具有一致性[10]. 但当光源的中心波长一定时, 随着谱宽的增大, 随机散射光学成像系统PSF的相关系数近似呈指数规律衰减, 并逐渐趋于零[15], 此时散斑光场的自相关信息与目标的自相关信息存在一定差异性. 通过(1)式可以得到宽谱散斑光场的自相关表示式为
式中, Imax和Imin分别为探测器接收到光强度最大与最小的图像, φ是光波场的初始相位. 由(6)式可知, 探测器接收到的图像强度随偏振方位角的改变呈现余弦函数变化规律. 为验证透过散射介质后形成的散斑光场具有偏振特性, 根据图3(a)搭建了宽谱散射成像实验装置. LED宽谱照明光源被第一个偏振器件调制为完全线偏振光, 依次照射至目标与散射介质后, 旋转第二个偏振调制器件的方位角, 每间隔5°采集序列宽谱散斑图像, 不同偏振方位角散斑图样的均值强度分布如图3(b)所示, 其中a, b, c, d分别为偏振方位角0°, 45°, 90°与135°的宽谱散斑图像. 图 3 宽谱散斑光场的偏振特性分析 (a)基于宽谱光源照明的偏振散射成像系统; (b)不同偏振方位角宽谱散斑图像的均值强度分布曲线 Figure3. Polarization characteristics analysis of broadband speckle field: (a) Polarization scattering imaging system with broadband light illumination; (b) the fitting curve between different rotated angles of polarizer and the mean intensity of broadband speckle.
根据不同偏振方位角散斑图样以及均值强度曲线可知, 散斑光场强度随偏振方位角有显著的明暗变化, 并且服从余弦函数分布规律. 为了深入分析散斑光场的偏振特性, 计算透过散射介质后散斑光场的偏振度[23], 其归一化强度分布如图4所示. 散斑光场中无论表征目标信息的散斑颗粒还是表征背景噪声信息的散斑图像都具有明显的偏振特性, 对于受背景噪声影响较小的散斑颗粒区域(如黄色圆圈所标记)其偏振度值大致位于0.3—0.9之间. 根据第2节的分析可知, 散斑光场中含有因光源谱宽引起的严重背景噪声项, 导致目标信息被噪声所淹没, 故而图4所示偏振度中所有的像素点取值取决于目标信息与背景噪声共同影响, 因此低于0.2值的像素点为背景噪声影响占优的结果. 以散斑偏振度图像的第300列(红色直线处)像素的归一化强度分布曲线为例, 偏振度的强度值在0.08至0.13之间波动, 变化趋势相对平稳且规律. 由以上分析可知, 图4表明基于偏振调制的宽谱照明光源透过散射介质形成的散斑光场中, 目标信息的偏振特性与大量背景噪声的偏振特性相比有显著的差异性[24]. 因此, 考虑充分利用散斑场中两者偏振特性的差异性提取出隐藏的目标信息, 从而从物理模型上解决宽谱散射成像中$ {C_2}\left( {{\lambda _{ij}}, \Delta {\lambda _{ij}}} \right){\kern 1 pt} $项的干扰问题. 图 4 宽谱散斑光场的偏振度归一化强度分布图 Figure4. The normalized intensity distribution of the degree of polarization of broadband speckle.
4.偏振散斑差分成像方法由于探测器所接收到的散斑光场在偏振域内呈现明显的偏振特性, 深入分析散斑光场中目标信息光和背景散射光的偏振差异性, 如图5所示. 通过旋转置于探测器之前的偏振片间隔5°采集序列散斑图像, 选取散斑中表征目标信息的2个散斑颗粒区域(图5(b)中的红色与绿色框)和背景信息的平滑区域(图5(b)中蓝色框), 进行像素强度值的统计, 区域强度分布曲线如图5(a)所示. 目标信息的变化随着偏振片的旋转方向总体呈现明显的余弦变化趋势, 表明其具有明显的偏振特性; 而背景散射信息的变化则相对比较稳定, 表明其呈现出弱偏振特性. 因此, 鉴于此差异性, 结合(7)式所示的偏振共模抑制特性方法有效滤除背景${C_2}\left(\right. {\lambda _{ij}}, $$ \Delta {\lambda _{ij}} \left.\right){\kern 1 pt}$项, 提取目标信息光. 图 5 不同偏振方位角散斑图像中目标和背景的强度变化 (a)目标与背景的强度分布曲线(O_1和O_2分别表示图(b)中表征目标信息的绿色和红色区域散斑颗粒强度分布, B为图(b)中表征背景信息的蓝色区域散斑颗粒强度分布); (b)不同偏振方位角散斑图像 (P1, P2, P3, ???, Pn表示探测器前偏振片在不同旋转方位角采集的散斑图样) Figure5. The intensity of the object and background as a function of different rotated angles of polarizer: (a) The fitting curves (O_1 and O_2 respectively represent the object information intensity distribution of speckle particles in the green and red regions of the figure (b), and B represents the background information intensity distribution of the speckle particles in the blue region of the figure (b)); (b) speckles with different rotated angles of polarizer (P1, P2, P3, ???, Pn represent the speckle patterns obtained by the polarizer in front of the detector at different rotated angles).