High-quality Reconstruction Research Articles

HDhuman: High-Quality Human Novel-View Rendering From Sparse Views.

In this paper, we aim to address the challenge of novel view rendering of human performers that wear clothes with complex texture patterns using a sparse set of camera views. Although some recent works have achieved remarkable rendering quality on humans with relatively uniform textures using sparse views, the rendering quality remains limited when dealing with complex texture patterns as they are unable to recover the high-frequency geometry details that are observed in the input views. To this end, we propose HDhuman, which uses a human reconstruction network with a pixel-aligned spatial transformer and a rendering network with geometry-guided pixel-wise feature integration to achieve high-quality human reconstruction and rendering. The designed pixel-aligned spatial transformer calculates the correlations between the input views and generates human reconstruction results with high-frequency details. Based on the surface reconstruction results, the geometry-guided pixel-wise visibility reasoning provides guidance for multi-view feature integration, enabling the rendering network to render high-quality images at 2k resolution on novel views. Unlike previous neural rendering works that always need to train or fine-tune an independent network for a different scene, our method is a general framework that is able to generalize to novel subjects. Experiments show that our approach outperforms all the prior generic or specific methods on both synthetic data and real-world data. Source code and test data will be made publicly available for research purposes at http://cic.tju.edu.cn/faculty/likun/projects/HDhuman/index.html.

Mechanical vibration signal compression based on speech codecs for intelligent manufacturing

ABSTRACT How to effectively compress mechanical signals so that they can support remote and real-time health monitoring is a hot issue in the context of intelligent manufacturing. Due to the characteristics of small repeatability, high noise, and high numerical accuracy of vibration signal, traditional compression algorithms cannot well balance the compression ratio, reconstruction quality, compression time indicators. Therefore, this paper presents a novel mechanical vibration signal compression scheme based on speech codecs, which can realize signal compression in many scenarios such as tool wear detection and bearing fault diagnosis. In this work, mechanical vibration signal compression is first transformed into a multi-objective optimization problem, and the feasibility of solving this problem from the perspective of speech coding is deeply analyzed. Further, this paper presents five possible compression methods with different compression ratios. The presented methods are evaluated with real-time data for tool wear detection in stone processing and public datasets for bearing fault diagnosis. The experimental results show that the proposed methods achieve high compression ratios (4.0 ~ 24.6) while ensuring low time overhead (the average time for compressing a sample point is not more than 0.0194 ms) and high reconstruction quality (the MSE index all less than 0.008), outperforming other state-of-the-art methods.

Image privacy protection scheme based on high-quality reconstruction DCT compression and nonlinear dynamics

The protection of digital image privacy in Big Data environment is a hot issue of increasing concern. To address the contradiction between efficiency and security for privacy protection, this paper proposes a high-quality restored image privacy protection scheme based on Discrete Cosine Transform (DCT) frequency domain compression and nonlinear dynamics. First, the RGB space of the color image is converted to YCbCr domain with 4:2:0 sampling. Then, the image is divided into sub-blocks in the spatial domain, followed by block-wise DCT and quantization into frequency domain information. Next, both Alternating Current (AC) and Direct Current (DC) coefficients of all sub-blocks are individually extracted and subjected to compression encoding. Finally, a Rubik’s Cube permutation and filtering diffusion encryption algorithm based on a nonlinear dynamical chaotic system is designed. Moreover, a mechanism for generating dynamic chaotic sequences associated with plaintext is introduced to effectively counter cryptographic attacks. The combined approach of compression and encryption significantly reduces computational complexity while improving encryption efficiency. Through experimental simulation results, the compression encryption scheme demonstrates high compression ratios as well as excellent image recovery quality while providing enhanced security against common cryptographic attacks. The work in this paper provides a preferred joint compression and encryption technical solution for digital image privacy protection in Big Data network.

PELMI: Realize robust DNA image storage under general errors via parity encoding and local mean iteration.

DNA molecules as storage media are characterized by high encoding density and low energy consumption, making DNA storage a highly promising storage method. However, DNA storage has shortcomings, especially when storing multimedia data, wherein image reconstruction fails when address errors occur, resulting in complete data loss. Therefore, we propose a parity encoding and local mean iteration (PELMI) scheme to achieve robust DNA storage of images. The proposed parity encoding scheme satisfies the common biochemical constraints of DNA sequences and the undesired motif content. It addresses varying pixel weights at different positions for binary data, thus optimizing the utilization of Reed-Solomon error correction. Then, through lost and erroneous sequences, data supplementation and local mean iteration are employed to enhance the robustness. The encoding results show that the undesired motif content is reduced by 23%-50% compared with the representative schemes, which improves the sequence stability. PELMI achieves image reconstruction under general errors (insertion, deletion, substitution) and enhances the DNA sequences quality. Especially under 1% error, compared with other advanced encoding schemes, the peak signal-to-noise ratio and the multiscale structure similarity address metric were increased by 10%-13% and 46.8%-122%, respectively, and the mean squared error decreased by 113%-127%. This demonstrates that the reconstructed images had better clarity, fidelity, and similarity in structure, texture, and detail. In summary, PELMI ensures robustness and stability of image storage in DNA and achieves relatively high-quality image reconstruction under general errors.

Temporal-Difference Graph-Based Optimization for High-Quality Reconstruction of MODIS NDVI Data

The Normalized Difference Vegetation Index (NDVI) is a crucial remote-sensing metric for assessing land surface vegetation greenness, essential for various studies encompassing phenology, ecology, hydrology, etc. However, effective applications of NDVI data are hindered by data noise due to factors such as cloud contamination, posing challenges for accurate observation. In this study, we proposed a novel approach for employing a Temporal-Difference Graph (TDG) method to reconstruct low-quality pixels in NDVI data. Regarding spatio-temporal NDVI data as a time-varying graph signal, the developed method utilized an optimization algorithm to maximize the spatial smoothness of temporal differences while preserving the spatial NDVI pattern. This approach was further evaluated by reconstructing MODIS/Terra Vegetation Indices 16-Day L3 Global 250 m Grid (MOD13Q1) products over Northwest China. Through quantitative comparison with a previous state-of-the-art method, the Savitzky–Golay (SG) filter method, the obtained results demonstrated the superior performance of the TDG method, and highly accurate results were achieved in both the temporal and spatial domains irrespective of noise types (positively-biased, negatively-biased, or linearly-interpolated noise). In addition, the TDG-based optimization approach shows great robustness to noise intensity within spatio-temporal NDVI data, suggesting promising prospects for its application to similar datasets.

Robust image tamper detection and recovery with self-embedding watermarking using SPIHT and LDPC

In today’s digital landscape, the pervasive use of digital images across diverse domains has led to growing concerns regarding their authenticity and reliability. The potential for malicious manipulation of these images underscores the critical need to develop robust methods for detecting tampering and ensuring their integrity. Fragile watermarking has been found to have extensive applications for tamper detection and recovery. An image watermarking technique for tamper detection, correction, and recovery is presented in this study. The proposed method employs a self-embedding method to generate the reference watermark from the original image, which has the advantage of superior tamper detection, localization, recovery capabilities, and robustness against attacks. Set Partitioning in the Hierarchical Tree (SPIHT) algorithm is applied to generate a reference watermark of the original image. Low-Density Parity Check (LDPC) is employed for error correction, providing higher-quality reconstruction to recover the original image. Schur decomposition processed the watermarked image blocks to generate authentication bits for each block to enhance tampering detection. The proposed watermarking method was evaluated using PSNR, SSIM, BER, and NC metrics for grayscale and colored images. The technique demonstrated high robustness in tamper detection and recovery against various malicious attacks. Comparative analysis with existing methods shows the efficacy of the proposed method.

Polarized-Speckle Deviation Imaging through Scattering Media under Strong Background Light Interference

A crucial challenge faced by noninvasive imaging through strongly scattering media is overcoming background light interference. Polarization-based anti-scattering methods can eliminate background light interference, but fail to utilize speckle images that do not contain unscattered object light for object reconstruction. Although speckle correlation imaging (SCI) methods can utilize speckle images for object reconstruction, it is difficult to achieve stable high-quality reconstruction and overcome background light interference using these methods. In this study, we propose a polarized-speckle deviation imaging (PSDI) method to overcome background light interference and achieve high-quality imaging through strongly scattering media. PSDI utilizes the bispectrum and autocorrelation of polarized speckle image deviations to reconstruct the Fourier phase and amplitude spectra of the object image, respectively. Experimental results show that when the background light is polarized and unpolarized, PSDI can achieve stable high-fidelity reconstruction of a polarized object when the signal-to-background ratio (SBR) is lower than −7 dB and −9 dB, respectively. PSDI bridges the gap between imaging with strongly scattered light and overcoming strong background light interference, and is expected to find widespread applications in fields such as biomedical imaging, astronomical observation, underwater imaging, and remote sensing.

Spatio-Temporal Evolution of Ecological Environment Quality Based on High-Quality Time-Series Data Reconstruction: A Case Study in the Sanjiangyuan Nature Reserve of China

The Sanjiangyuan Nature Reserve of China (SNRC) is recognized as one of the most fragile and sensitive terrestrial ecosystems in China, posing challenges for obtaining reliable and complete Moderate Resolution Imaging Spectro Radiometer (MODIS) data for ecological environment quality (EEQ) monitoring due to adverse factors like clouds and snow. In this study, a complete high-quality framework for MODIS time-series data reconstruction was constructed utilizing the Google Earth Engine (GEE) cloud platform. The reconstructed images were used to compute the Remote Sensing based Ecological Index (RSEI) on a monthly scale in the SNRC from 2001 to 2020. The results were as follows: The EEQ of the study area exhibited a “first fluctuating decline, then significant improvement” trend, with the RSEI values increasing at a rate of 0.84%/a. The spatial pattern of the EEQ displayed significant spatial heterogeneity, characterized by a “low in the west and high in the east” distribution. The spatial distribution pattern of the RSEI exhibited significant clustering characteristics. From 2001 to 2020, the proportion of “high–high” clustering areas exceeded 35%, and the proportion of “low–low” clustering areas exceeded 30%. Poor ecological conditions are mainly associated with population agglomerations, cultivated land, unutilized land, and bare ground, while grasslands and forests have higher RSEI values. The result of the trend analysis revealed a significant trend in RSEI change, with 62.96% of the area significantly improved and 6.31% significantly degraded. The Hurst Index (HI) results indicated that the future trend of the RSEI is predominantly anti-persistence. The proportion of areas where the EEQ is expected to continue improving in the future is 33.74%, whereas 21.21% of the area is forecasted to transition from improvement to degradation. The results showed that the high-quality framework for MODIS time-series data reconstruction enables the effective continuous monitoring of EEQ over long periods and large areas, providing robust scientific support for long time-series data reconstruction research.

Q2-metnet: QIIME2 package to analyse 16S rRNA data via high-quality metabolic reconstructions of the human gut microbiota.

16S rRNA gene sequencing is the most frequent approach for the characterization of the human gut microbiota. Despite different efforts in the literature, the inference of functional and metabolic interpretations from 16S rRNA gene sequencing data is still a challenging task. High-quality metabolic reconstructions of the human gut microbiota, such as AGORA and AGREDA, constitute a curated resource to improve functional inference from 16S rRNA data, but they are not typically integrated into standard bioinformatics tools. Here, we present q2-metnet, a QIIME2 plugin that enables the contextualization of 16S rRNA gene sequencing data into AGORA and AGREDA. In particular, based on relative abundances of taxa, q2-metnet determines normalized activity scores for the reactions and subsystems involved in the selected metabolic reconstruction. Using these scores, q2-metnet allows the user to conduct differential activity analysis for reactions and subsystems, as well as exploratory analysis using PCA and hierarchical clustering. We apply q2-metnet to a dataset from our group that involves 16S rRNA data from stool samples from lean, allergic to cow's milk, obese and celiac children, and the Belgian Flemish Gut Flora Project cohort, which includes faecal 16S rRNA data from obese and normal-weight adult individuals. In the first case, q2-metnet outperforms existing algorithms in separating different clinical conditions based on predicted pathway abundances and subsystem scores. In the second case, q2-metnet complements competing approaches in predicting functional alterations in the gut microbiota of obese individuals. Overall, q2-metnet constitutes a powerful bioinformatics tool to provide metabolic context to 16S rRNA data from the human gut microbiota. Python code of q2-metnet is available in https://github.com/PlanesLab/q2-metnet and https://figshare.com/articles/dataset/q2-metnet_package/26180446.

Explicit 3D reconstruction from images with dynamic graph learning and rendering-guided diffusion

High-quality 3D reconstruction is becoming increasingly important in a variety of fields. Recently, implicit representation methods have made significant progress in image-based 3D reconstruction. However, these methods tend to yield entangled neural representations which lack support for standard 3D pipelines, and their reconstruction results usually experience a sharp drop in quality when input views are reduced. To obtain high-quality 3D content, we propose an explicit 3D reconstruction method that directly extracts textured meshes from images and remains robust using reduced input views. Our central components include a dynamic graph convolutional network (GCN) and a rendering-guided diffusion model. The dynamic GCN aims to improve mesh reconstruction quality by effectively aggregating features from vertex neighborhoods. The aggregation is accelerated through sampling geometric-related neighbors with different SDF signs, which gradually converges in quantity during training. The rendering-guided diffusion model learns prior distributions for unseen regions to improve reconstruction performance using sparse-view inputs. It uses the rendered image under an interpolated camera pose as conditioned input and its diffusion strength can be controlled with the rendering loss of explicit reconstruction. In addition, the rendering-guided diffusion model can be jointly trained to generate plausible novel views with 3D consistency. Experiments demonstrate that our method can produce high-quality explicit reconstruction results and maintain realistic reconstruction using sparse-view inputs.

IW-NeRF: Using Implicit Watermarks to Protect the Copyright of Neural Radiation Fields

The neural radiance field (NeRF) has demonstrated significant advancements in computer vision. However, the training process for NeRF models necessitates extensive computational resources and ample training data. In the event of unauthorized usage or theft of the model, substantial losses can be incurred by the copyright holder. To address this concern, we present a novel algorithm that leverages the implicit neural representation (INR) watermarking technique to safeguard NeRF model copyrights. By encoding the watermark information implicitly, we integrate its parameters into the NeRF model’s network using a unique key. Through this key, the copyright owner can extract the embedded watermarks from the NeRF model for ownership verification. To the best of our knowledge, this is the pioneering implementation of INR watermarking for the protection of NeRF model copyrights. Our experimental results substantiate that our approach not only offers robustness and preserves high-quality 3D reconstructions but also ensures the flawless (100%) extraction of watermark content, thereby effectively securing the copyright of the NeRF model.

Calculating point spread functions: methods, pitfalls, and solutions.

The knowledge of the exact structure of the optical system point spread function (PSF) enables a high-quality image reconstruction in fluorescence microscopy. Accurate PSF models account for the vector nature of light and the phase and amplitude modifications. Most existing real-space-based PSF models fall into a sampling pitfall near the center position, yielding to the violation of energy conservation. In this work, we present a novel, to the best of our knowledge, Fourier-based techniques for computing vector PSF and compare them to the state-of-the-art. Our methods are shown to satisfy the physical condition of the imaging process. They are reproducible, computationally efficient, easy to implement, and easy to modify to represent various imaging modalities.

Reconstruction of dense time series high spatial resolution NDVI data using a spatiotemporal optimal weighted combination estimation model based on Sentinel-2 and MODIS

A regular, dense time series of the normalized difference vegetation index (NDVI) is a crucial remote sensing parameter that provides insights into the growth status of plants across different temporal and spatial scales. However, the production of high-resolution, equal-interval, and dense-time-series NDVI products currently faces technical challenges, limiting their application in agriculture and forestry-related fields. This study proposes a new method for reconstructing a dense Sentinel-2 NDVI time series based on the spatiotemporal optimal weighted combination estimation model (SOWCEM). The SOWCEM is developed using state-of-the-art spatial and temporal reconstruction algorithms. This model considers the distribution characteristics of errors in spatiotemporal dimensions, allowing for adaptive determination of the optimal combination weight at the pixel level. Then, the reconstructed regular dense-time-series NDVI data covering the region for five days/time are evaluated through a comparative analysis. The experimental results demonstrate that the regional NDVI reconstruction image based on the SOWCEM algorithm has a determination coefficient (R2) of 0.9082, a root mean square error (RMSE) of 0.0403, and a comprehensive overall error (ERGAS) of 5.7726. Compared with single spatiotemporal dimension models, the SOWCEM algorithm shows an average improvement of 5.78% in R2, while the RMSE and ERGAS decrease by 6.5% and 6.86%, respectively. Moreover, the stability and robustness of the models are also enhanced. These results indicate that the SOWCEM algorithm can effectively achieve a high-quality fusion reconstruction of regional images, significantly enhancing the accuracy of high-spatial-resolution NDVI dense temporal spectral reconstruction. Furthermore, it exhibits clear superiority and good applicability. This algorithmic framework provides a feasible approach for the dense reconstruction of high-spatial-resolution regular time-series NDVI data, enabling more accurate and efficient high-resolution NDVI remote sensing monitoring services in related field applications. The core code of SOWCEM is available at https://github.com/GISerZkun/SOWCEM.

LD-CSNet: A latent diffusion-based architecture for perceptual Compressed Sensing

Compressed Sensing (CS) is a groundbreaking paradigm in image acquisition, challenging the constraints of the Nyquist–Shannon sampling theorem. This enables high-quality image reconstruction using a minimal number of measurements. Neural Networks’ potent feature induction capabilities enable advanced data-driven CS methods to achieve high-fidelity image reconstruction. However, achieving satisfactory reconstruction performance, particularly in terms of perceptual quality, remains challenging at extremely low sampling rates. To tackle this challenge, we introduce a novel two-stage image CS framework based on latent diffusion, named LD-CSNet. In the first stage, we utilize an autoencoder pre-trained on a large dataset to represent natural images as low-dimensional latent vectors, establishing prior knowledge distinct from sparsity and effectively reducing the dimensionality of the solution space. In the second stage, we employ a conditional diffusion model for maximum likelihood estimates in the latent space. This is supported by a measurement embedding module designed to encode measurements, making them suitable for a denoising network. This guides the generation process in reconstructing low-dimensional latent vectors. Finally, the image is reconstructed using a pre-trained decoder. Experimental results across multiple public datasets demonstrate LD-CSNet’s superior perceptual quality and robustness to noise. It maintains fidelity and visual quality at lower sampling rates. Research findings suggest the promising application of diffusion models in image CS. Future research can focus on developing more appropriate models for the first stage.

Reconstruction of OFDM Signals Using a Dual Discriminator CGAN with BiLSTM and Transformer.

Communication signal reconstruction technology represents a critical area of research within communication countermeasures and signal processing. Considering traditional OFDM signal reconstruction methods' intricacy and suboptimal reconstruction performance, a dual discriminator CGAN model incorporating LSTM and Transformer is proposed. When reconstructing OFDM signals using the traditional CNN network, it becomes challenging to extract intricate temporal information. Therefore, the BiLSTM network is incorporated into the first discriminator to capture timing details of the IQ (In-phase and Quadrature-phase) sequence and constellation map information of the AP (Amplitude and Phase) sequence. Subsequently, following the addition of fixed position coding, these data are fed into the core network constructed based on the Transformer Encoder for further learning. Simultaneously, to capture the correlation between the two IQ signals, the VIT (Vision in Transformer) concept is incorporated into the second discriminator. The IQ sequence is treated as a single-channel two-dimensional image and segmented into pixel blocks containing IQ sequence through Conv2d. Fixed position coding is added and sent to the Transformer core network for learning. The generator network transforms input noise data into a dimensional space aligned with the IQ signal and embedding vector dimensions. It appends identical position encoding information to the IQ sequence before sending it to the Transformer network. The experimental results demonstrate that, under commonly utilized OFDM modulation formats such as BPSK, QPSK, and 16QAM, the time series waveform, constellation diagram, and spectral diagram exhibit high-quality reconstruction. Our algorithm achieves improved signal quality while managing complexity compared to other reconstruction methods.

The Strategy of Digital twin Implementation at the Cyber-Physical Enterprises

The strategy of digital twin (DT) implementation at the cyber-physical enterprises is proposed, which is due to the need for high-quality and effective reconstruction of Ukraine. The TISM and MICMAC methods are used to study the factors that influence the strategy of DT implementation to support the guaranteed functioning of a cyber-physical system. 13 main factors, that determine strategy success, were identified, and a model of strategy realization was built in the form of a six-level interaction diagram (digraph). The factors are classified by influence and dependence. The substantiated conclusion is that the state support of DT implementation strategy, the availability of design standards, data protection methods, legislative regulation and implementation experience by other enterprises are key factors to the success of strategy for DT implementation at the enterprise.

All-dielectric metasurface for polarization-multiplexed single-pixel imaging

Integration and miniaturization of multi-channel single-pixel imaging systems have become a developing trend. However, it is challenging to meet such development needs solely relying on traditional optical devices. One feasible solution is the utilization of metasurfaces with multiplexing functionality. Here, we propose and validate an all-dielectric, anisotropic metasurface that provides a random mask with polarization multiplexing for single pixel imaging. The design ensures each mask contains 50% target information, allowing adaptive correlated imaging of different targets without needing to redesign the masks. By optimizing the metasurface, we enhance computational efficiency by preventing correlation between different polarization channels and mask patterns. We also adjust the parameters of the compressed sensing algorithm to accommodate various sampling rates, ensuring high-quality image reconstruction. Additionally, the whole system is simulated by the angular spectrum transmission and compressed sensing reconstruction algorithm, providing image reconstruction results for metasurfaces of different sizes, demonstrating the feasibility of the proposed approach. It is noteworthy that the designed metasurface works for single-wavelength operation and could be extended to multispectral imaging by introducing achromatic metasurface technology. The proposed method could miniaturize the optical devices and reduce light loss.

A machine learning decision criterion for reducing scan time for hyperspectral neutron computed tomography systems

We present the first machine learning-based autonomous hyperspectral neutron computed tomography experiment performed at the Spallation Neutron Source. Hyperspectral neutron computed tomography allows the characterization of samples by enabling the reconstruction of crystallographic information and elemental/isotopic composition of objects relevant to materials science. High quality reconstructions using traditional algorithms such as the filtered back projection require a high signal-to-noise ratio across a wide wavelength range combined with a large number of projections. This results in scan times of several days to acquire hundreds of hyperspectral projections, during which end users have minimal feedback. To address these challenges, a golden ratio scanning protocol combined with model-based image reconstruction algorithms have been proposed. This novel approach enables high quality real-time reconstructions from streaming experimental data, thus providing feedback to users, while requiring fewer yet a fixed number of projections compared to the filtered back projection method. In this paper, we propose a novel machine learning criterion that can terminate a streaming neutron tomography scan once sufficient information is obtained based on the current set of measurements. Our decision criterion uses a quality score which combines a reference-free image quality metric computed using a pre-trained deep neural network with a metric that measures differences between consecutive reconstructions. The results show that our method can reduce the measurement time by approximately a factor of five compared to a baseline method based on filtered back projection for the samples we studied while automatically terminating the scans.

Approaching Maximum Resolution in Structured Illumination Microscopy via Accurate Noise Modeling.

Biological images captured by microscopes are characterized by heterogeneous signal-to-noise ratios (SNRs) due to spatially varying photon emission across the field of view convoluted with camera noise. State-of-the-art unsupervised structured illumination microscopy (SIM) reconstruction algorithms, commonly implemented in the Fourier domain, do not accurately model this noise and suffer from high-frequency artifacts, user-dependent choices of smoothness constraints making assumptions on biological features, and unphysical negative values in the recovered fluorescence intensity map. On the other hand, supervised methods rely on large datasets for training, and often require retraining for new sample structures. Consequently, achieving high contrast near the maximum theoretical resolution in an unsupervised, physically principled, manner remains an open problem. Here, we propose Bayesian-SIM (B-SIM), an unsupervised Bayesian framework to quantitatively reconstruct SIM data, rectifying these shortcomings by accurately incorporating known noise sources in the spatial domain. To accelerate the reconstruction process, we use the finite extent of the point-spread-function to devise a parallelized Monte Carlo strategy involving chunking and restitching of the inferred fluorescence intensity. We benchmark our framework on both simulated and experimental images, and demonstrate improved contrast permitting feature recovery at up to 25% shorter length scales over state-of-the-art methods at both high- and low-SNR. B-SIM enables unsupervised, quantitative, physically accurate reconstruction without the need for labeled training data, democratizing high-quality SIM reconstruction and expands the capabilities of live-cell SIM to lower SNR, potentially revealing biological features in previously inaccessible regimes.

Polarimetric Helmholtz Stereopsis.

Helmholtz stereopsis (HS) exploits the reciprocity principle of light propagation (i.e., the Helmholtz reciprocity) for 3D reconstruction of surfaces with arbitrary reflectance. In this paper, we present the polarimetric Helmholtz stereopsis (polar-HS), which extends the classical HS by considering the polarization state of light in the reciprocal paths. With the additional phase information from polarization, polar-HS requires only one reciprocal image pair. We derive the reciprocity relationship of Mueller matrix and formulate new reciprocity constraint that takes polarization state into account. We also utilize polarimetric constraints and extend them to the case of perspective projection. For the recovery of surface depths and normals, we incorporate reciprocity constraint with diffuse/specular polarimetric constraints in a unified optimization framework. For depth estimation, we further propose to utilize the consistency of diffuse angle of polarization. For normal estimation, we develop a normal refinement strategy based on degree of linear polarization. Using a hardware prototype, we show that our approach produces high-quality 3D reconstruction for different types of surfaces, ranging from diffuse to highly specular.