Organizations: School of Electrical Engineering, The Iby and Aladar Fleischman Faculty of Engineering, Tel Aviv University, Tel Aviv, Israel · Current affiliation: Department of Physics of Complex Systems, Faculty of Physics, Weizmann Institute of Science, Rehovot, Israel. · Triangle Regional Research and Development Center, Kfar Qara’ 3007500, Israel
Polarization imaging is a technique that creates a pixel map of the polarization state in a scene. Although invisible to the human eye, polarization can assist various sensing and computer vision tasks. Existing polarization cameras use spatial or temporal multiplexing, which increases the camera volume, weight, cost, or all of the above. Recent lensless imaging approaches, such as DiffuserCam, have demonstrated that compact imaging systems can be realized by replacing the lens with a coding element and performing computational reconstruction. In this work, we propose a compact lensless polarization camera composed of a diffuser and a simple striped polarization mask. By combining this optical design with a reconstruction algorithm that explicitly models the polarization-encoded lensless measurements, four linear polarization images are recovered from a single snapshot. Our results demonstrate the potential of lensless approaches for polarization imaging and reveal the physical factors that govern reconstruction quality, guiding the development of high-quality practical systems.
Figures & tables
Figure 1: Overview of the proposed lensless polarization camera Given (a) a scene with polarized light, the imaging is performed using (b) a random diffuser. Polarization multiplexing is achieved using (c) a mask located on the sensor plane, where (e) demonstrates a single period of the polarization multiplexing pattern, encoding linear polarization orientations ( 0∘,45∘,90∘ and 135∘ ). This imaging scheme results in (d) a diffused and multiplexed image. Using the diffuser PSF and polarization mask characteristics, the reconstruction algorithm recovers (f) polarization intensity images of the scene.
Figure 2: Optical encoding elements. (a) Diffuser PSF. The random structure of the diffuser acts as multiple randomly distributed impulse-like features. This produces a spatially extended PSF with many narrow features that enable image reconstruction. (b) Polarization mask. Leveraging the ability to reconstruct images from partial sampling of the diffuser-encoded measurement, the mask is composed of linear polarizer stripes at the required orientations. The mask response is measured under back-illumination using a polarized light source, enabling the different transmissions to be clearly visualized. (c) A prototype lensless polarization camera integrating the diffuser and polarization mask at the sensor.
Figure 3: Front illumination — matched-forward-model simulation. For each scene, the first row shows the lensless polarization reconstruction with the same polarization mask for data generation and reconstruction, and the second row shows the corresponding spatially aligned reference images acquired using a lens-based camera with an external rotating polarizer.
Figure 4: Back illumination - matched-forward-model simulation: (top) lensless reconstruction using a measured polarization mask (matched setting),(middle) lensless reconstruction using an idealized polarization mask (matched setting), and (bottom) corresponding spatially aligned reference images acquired using a lens-based camera with an external rotating polarizer.
Figure 5: Measured and perturbed polarization mask models used in the mismatch analysis. Top row: measured polarization mask responses corresponding to the four polarization orientations ( 0∘,45∘,90∘,135∘ ) obtained from the fabricated mask. Bottom row: examples of perturbed mask models derived from the measured 0∘ mask, including additive Gaussian noise, Gaussian blurring, interpolation with an idealized simulated mask (mean), and an idealized simulated mask.
Figure 6: Robustness of lensless polarization reconstruction to polarization-mask model mismatch. Top: Gaussian blurring of the measured mask ( σ=0 corresponds to the unblurred measured mask used for both data generation and reconstruction). Middle: additive Gaussian noise applied to the simulated mask ( σ=0 corresponds to the noise-free simulated mask used for both data generation and reconstruction). Bottom: interpolation between an idealized simulated mask and the measured physical mask, with t=a+bb ( t=0 : normalized measured mask used for both data generation and reconstruction; t=1 : simulated mask used for both). Left: PSNR. Right: SSIM. Solid curves show the mean across the scenes shown in Figures 3 and 4 with ±1 standard deviation shading; dashed curves denote individual scenes.
Figure 7: Front illumination example. (top) Lensless polarization reconstruction from a single snapshot acquired with the lab prototype. (bottom) Reconstruction of matched-mask setting obtained by synthetically multiplexing polarization lensless measurements using the measured polarization mask. The real and simulated scenes are similar but not identical.
Appendix figures & tables4 assets
Supplementary material from the paper’s appendix.
Appendix
Figure 8: Effect of polarization mask mismatch on reconstruction quality. Top row: reference lensless reconstructions for the I0 polarization channel obtained (i) without a polarization mask (using a diffuser and external polarizer), (ii) under a matched-mask setting in which the measured polarization mask is used consistently for both data generation and reconstruction, and (iii) using the physical polarization mask in the prototype. Bottom row: reconstructions obtained under controlled polarization-mask mismatch (1D spatial mask blur, additive mask noise, and interpolation between the ideal and measured masks), following the protocol described in Section 4.2 of the main paper.
Figure 9: Effect of mask-sensor distance on the system PSF. (Left) Sensor-plane intensity without the polarization mask - a diffraction-limited PSF pattern. (Middle) Sensor-plane intensity with the polarization mask placed 1.66mm in front of the sensor, showing pronounced spreading perpendicular to the stripe direction. (Right) Center-line normalized intensity profiles extracted along the horizontal axis for both cases.
Figure 10: Back illumination examples For each scene, the first row shows the lensless polarization reconstruction using the physical prototype. The second row shows the corresponding reconstruction in the matched-mask simulation setting, where the measured mask is used for simulation and reconstruction. Columns correspond to the polarization sub-images I0 , I45 , I90 , and I135 .
Figure 11: Front illumination examples For each scene, the first row shows the lensless polarization reconstruction using the physical prototype. The second row shows the corresponding reconstruction in the matched-mask simulation setting, where the measured mask is used for simulation and reconstruction. Columns correspond to the polarization sub-images I0 , I45 , I90 , and I135 .
Polarization imaging captures the polarization state of light, revealing information invisible to the human eye yet valuable in domains such as biomedical diagnostics, autonomous driving, and remote sensing. However, conventional polarization cameras are often expensive, bulky, or both, limiting their practical use. Lensless imaging offers a compact, low-cost alternative by replacing the lens with a simple optical element like a diffuser and performing computational reconstruction, but existing lensless polarization systems suffer from limited reconstruction quality. To overcome these limitations, we introduce a RGB-guided lensless polarization imaging system that combines a compact polarization-RGB sensor with an auxiliary, widely available conventional RGB camera providing structural guidance. We reconstruct multi-angle polarization images for each RGB color channel through a two-stage pipeline: a physics-based inversion recovers an initial polarization image, followed by a Transformer-based fusion network that refines this reconstruction using the RGB guidance image from the conventional RGB camera. Our two-stage method significantly improves reconstruction quality and fidelity over lensless-only baselines, generalizes across datasets and imaging conditions, and achieves high-quality real-world results on our physical prototype lensless camera without any fine-tuning.
Noa Kraicer, Erez Yosef, Raja Giryes
Tel Aviv University School of Electrical Engineering, Faculty of Engineering
Polarization-based vision has gained increasing attention for providing richer physical cues beyond RGB images. While achieving single-shot capture is highly desirable for practical applications, existing Division-of-Focal-Plane (DoFP) sensors inherently suffer from reduced spatial resolution and artifacts due to their spatial multiplexing mechanism. To overcome these limitations without sacrificing the snapshot capability, we propose EasyPolar, a multi-view polarimetric imaging framework. Our system is grounded in the physical insight that three independent intensity measurements are sufficient to fully characterize linear polarization. Guided by this, we design a triple-camera setup consisting of three synchronized RGB cameras that capture one unpolarized view and two polarized views with distinct orientations. Building upon this hardware design, we further propose a confidence-guided polarization reconstruction network to address the potential misalignment in multi-view fusion. The network performs multi-modal feature fusion under a confidence-aware physical guidance mechanism, which effectively suppresses warping-induced artifacts and enforces explicit geometric constraints on the solution space. Experimental results demonstrate that our method achieves high-quality results and benefits various downstream tasks.
Shuangfan Zhou, Chu Zhou, Heng Guo +4
School of Artificial Intelligence, Beijing University of Posts and Telecommunications, China · National Institute of Informatics, Japan · State Key Laboratory for Multimedia Information Processing, School of Computer Science, Peking University, China +1
Polarimetric imaging enables advanced vision applications such as normal estimation and de-reflection by capturing unique surface-material interactions. However, existing applications (alternatively called downstream tasks) rely on datasets constructed by naively regrouping raw measurements from division-of-focal-plane sensors, where pixels of the same polarization angle are extracted and aligned into sparse images without proper demosaicking. This reconstruction strategy results in suboptimal, incomplete targets that limit downstream performance. Moreover, current demosaicking methods are task-agnostic, optimizing only for photometric fidelity rather than utility in downstream tasks. Towards this end, we propose PolarAPP, the first framework to jointly optimize demosaicking and its downstream tasks. PolarAPP introduces a feature alignment mechanism that semantically aligns the representations of demosaicking and downstream networks via meta-learning, guiding the reconstruction to be task-aware. It further employs an equivalent imaging constraint for demosaicking training, enabling direct regression to physically meaningful outputs without relying on rearranged data. Finally, a task-refinement stage fine-tunes the task network using the stable demosaicking front-end to further enhance accuracy. Extensive experimental results demonstrate that PolarAPP outperforms existing methods in both demosaicking quality and downstream performance. Code is available upon acceptance.
Yidong Luo, Chenggong Li, Yunfeng Song +4
Zhejiang University, Hangzhou, China · School of Automation, Central South University, Changsha, China · School of Engineering, Westlake University, Hangzhou, China +1