Neural rendering with implicit neural networks has recently emerged as an attractive proposition for scene reconstruction, achieving excellent quality albeit at high computational cost. While the most recent generation of such methods has made progress on the rendering (inference) times, very little progress has been made on improving the reconstruction (training) times. In this work, we present Neural Adaptive Scene Tracing (NAScenT), the first neural rendering method based on directly training a hybrid explicit-implicit neural representation. NAScenT uses a hierarchical octree representation with one neural network per leaf node and combines this representation with a two-stage sampling process that concentrates ray samples where they matter most near object surfaces. As a result, NAScenT is capable of reconstructing challenging scenes including both large, sparsely populated volumes like UAV captured outdoor environments, as well as small scenes with high geometric complexity. NAScenT outperforms existing neural rendering approaches in terms of both quality and training time.
In this paper, we present Neural Adaptive Tomography (NeAT), the first adaptive, hierarchical neural rendering pipeline for multi-view inverse rendering. Through a combination of neural features with an adaptive explicit representation, we achieve reconstruction times far superior to existing neural inverse rendering methods. The adaptive explicit representation improves efficiency by facilitating empty space culling and concentrating samples in complex regions, while the neural features act as a neural regularizer for the 3D reconstruction. The NeAT framework is designed specifically for the tomographic setting, which consists only of semi-transparent volumetric scenes instead of opaque objects. In this setting, NeAT outperforms the quality of existing optimization-based tomography solvers while being substantially faster.
This paper study the reconstruction of High Dynamic Range (HDR) video from snapshot-coded LDR video. Constructing an HDR video requires restoring the HDR values for each frame and maintaining the consistency between successive frames. HDR image acquisition from single image capture, also known as snapshot HDR imaging, can be achieved in several ways. For example, the reconfigurable snapshot HDR camera is realized by introducing an optical element into the optical stack of the camera; by placing a coded mask at a small standoff distance in front of the sensor. High-quality HDR image can be recovered from the captured coded image using deep learning methods. This study utilizes 3D-CNNs to perform a joint demosaicking, denoising, and HDR video reconstruction from coded LDR video. We enforce more temporally consistent HDR video reconstruction by introducing a temporal loss function that considers the short-term and long-term consistency. The obtained results are promising and could lead to affordable HDR video capture using conventional cameras.
Under-display cameras have been proposed in recent years as a way to reduce the form factor of mobile devices while maximizing the screen area. Unfortunately, placing the camera behind the screen results in significant image distortions, including loss of contrast, blur, noise, color shift, scattering artifacts, and reduced light sensitivity. In this paper, we propose an image-restoration pipeline that is ISP-agnostic, i.e. it can be combined with any legacy ISP to produce a final image that matches the appearance of regular cameras using the same ISP. This is achieved with a deep learning approach that performs a RAW-to-RAW image restoration. To obtain large quantities of real under-display camera training data with sufficient contrast and scene diversity, we furthermore develop a data capture method utilizing an HDR monitor, as well as a data augmentation method to generate suitable HDR content. The monitor data is supplemented with real-world data that has less scene diversity but allows us to achieve fine detail recovery without being limited by the monitor resolution. Together, this approach successfully restores color and contrast as well as image detail.
Simultaneous reconstruction of geometry and reflectance properties in uncontrolled environments remains a challenging problem. In this paper, we propose an efficient method to reconstruct the scene's 3D geometry and reflectance from multi-view photography using conventional hand-held cameras. Our method automatically builds a virtual scene in a differentiable rendering system that roughly matches the real world's scene parameters, optimized by minimizing photometric objectives alternatingly and stochastically. With the optimal scene parameters evaluated, photo-realistic novel views for various viewing angles and distances can then be generated by our approach. We present the results of captured scenes with complex geometry and various reflection types. Our method also shows superior performance compared to state-of-the-art alternatives in novel view synthesis visually and quantitatively.
Lensless cameras are a class of imaging devices that shrink the physical dimensions to the very close vicinity of the image sensor by integrating flat optics and computational algorithms. Here we report a flat lensless camera with spatially-coded Voronoi-Fresnel phase, partly inspired by biological apposition compound eye, to achieve superior image quality. We propose a design principle of maximizing the information in optics to facilitate the computational reconstruction. By introducing a Fourier domain metric, Modulation Transfer Function volume (MTFv), we devise an optimization framework to guide the optimal design of the optical element. The resulting Voronoi-Fresnel phase features an irregular array of quasi-Centroidal Voronoi cells containing a base first-order Fresnel phase function. We demonstrate and verify the imaging performance with a prototype Voronoi-Fresnel lensless camera on a 1.6-megapixel image sensor in various illumination conditions. The proposed design could benefit the development of compact imaging systems working in extreme physical conditions.
Holographic displays can generate light fields by dynamically modulating the wavefront of a coherent beam of light using a spatial light modulator, promising rich virtual and augmented reality applications. However, the limited spatial resolution of existing dynamic spatial light modulators imposes a tight bound on the diffraction angle. As a result, today's holographic displays possess low \'{e}tendue, which is the product of the display area and the maximum solid angle of diffracted light. The low \'{e}tendue forces a sacrifice of either the field of view (FOV) or the display size. In this work, we lift this limitation by presenting neural \'{e}tendue expanders. This new breed of optical elements, which is learned from a natural image dataset, enables higher diffraction angles for ultra-wide FOV while maintaining both a compact form factor and the fidelity of displayed contents to human viewers. With neural \'{e}tendue expanders, we achieve 64$\times$ \'{e}tendue expansion of natural images with reconstruction quality (measured in PSNR) over 29dB on simulated retinal-resolution images. As a result, the proposed approach with expansion factor 64$\times$ enables high-fidelity ultra-wide-angle holographic projection of natural images using an 8K-pixel SLM, resulting in a 18.5 mm eyebox size and 2.18 steradians FOV, covering 85\% of the human stereo FOV.
We introduce Mask-ToF, a method to reduce flying pixels (FP) in time-of-flight (ToF) depth captures. FPs are pervasive artifacts which occur around depth edges, where light paths from both an object and its background are integrated over the aperture. This light mixes at a sensor pixel to produce erroneous depth estimates, which can adversely affect downstream 3D vision tasks. Mask-ToF starts at the source of these FPs, learning a microlens-level occlusion mask which effectively creates a custom-shaped sub-aperture for each sensor pixel. This modulates the selection of foreground and background light mixtures on a per-pixel basis and thereby encodes scene geometric information directly into the ToF measurements. We develop a differentiable ToF simulator to jointly train a convolutional neural network to decode this information and produce high-fidelity, low-FP depth reconstructions. We test the effectiveness of Mask-ToF on a simulated light field dataset and validate the method with an experimental prototype. To this end, we manufacture the learned amplitude mask and design an optical relay system to virtually place it on a high-resolution ToF sensor. We find that Mask-ToF generalizes well to real data without retraining, cutting FP counts in half.
To extend the capabilities of spectral imaging, hyperspectral and depth imaging have been combined to capture the higher-dimensional visual information. However, the form factor of the combined imaging systems increases, limiting the applicability of this new technology. In this work, we propose a monocular imaging system for simultaneously capturing hyperspectral-depth (HS-D) scene information with an optimized diffractive optical element (DOE). In the training phase, this DOE is optimized jointly with a convolutional neural network to estimate HS-D data from a snapshot input. To study natural image statistics of this high-dimensional visual data and to enable such a machine learning-based DOE training procedure, we record two HS-D datasets. One is used for end-to-end optimization in deep optical HS-D imaging, and the other is used for enhancing reconstruction performance with a real-DOE prototype. The optimized DOE is fabricated with a grayscale lithography process and inserted into a portable HS-D camera prototype, which is shown to robustly capture HS-D information. In extensive evaluations, we demonstrate that our deep optical imaging system achieves state-of-the-art results for HS-D imaging and that the optimized DOE outperforms alternative optical designs.
State-of-the-art methods for Convolutional Sparse Coding usually employ Fourier-domain solvers in order to speed up the convolution operators. However, this approach is not without shortcomings. For example, Fourier-domain representations implicitly assume circular boundary conditions and make it hard to fully exploit the sparsity of the problem as well as the small spatial support of the filters. In this work, we propose a novel stochastic spatial-domain solver, in which a randomized subsampling strategy is introduced during the learning sparse codes. Afterwards, we extend the proposed strategy in conjunction with online learning, scaling the CSC model up to very large sample sizes. In both cases, we show experimentally that the proposed subsampling strategy, with a reasonable selection of the subsampling rate, outperforms the state-of-the-art frequency-domain solvers in terms of execution time without losing the learning quality. Finally, we evaluate the effectiveness of the over-complete dictionary learned from large-scale datasets, which demonstrates an improved sparse representation of the natural images on account of more abundant learned image features.