Search

Image distortion by atmospheric turbulence is a stochastic degradation, which is a critical problem in long-range optical imaging systems. A number of research has been conducted during the past decades, including model-based and emerging deep-learning solutions with the help of synthetic data. Although fast and physics-grounded simulation tools have been introduced to help the deep-learning models adapt to real-world turbulence conditions recently, the training of such models only relies on the synthetic data and ground truth pairs. This paper proposes the Physics-integrated Restoration Network (PiRN) to bring the physics-based simulator directly into the training process to help the network to disentangle the stochasticity from the degradation and the underlying image. Furthermore, to overcome the ``average effect" introduced by deterministic models and the domain gap between the synthetic and real-world degradation, we further introduce PiRN with Stochastic Refinement (PiRN-SR) to boost its perceptual quality. Overall, our PiRN and PiRN-SR improve the generalization to real-world unknown turbulence conditions and provide a state-of-the-art restoration in both pixel-wise accuracy and perceptual quality. Our codes are available at \url{https://github.com/VITA-Group/PiRN}., Accepted by ICCV 2023

The advancement of new digital image sensors has enabled the design of exposure multiplexing schemes where a single image capture can have multiple exposures and conversion gains in an interlaced format, similar to that of a Bayer color filter array. In this paper, we ask the question of how to design such multiplexing schemes for adaptive high-dynamic range (HDR) imaging where the multiplexing scheme can be updated according to the scenes. We present two new findings. (i) We address the problem of design optimality. We show that given a multiplex pattern, the conventional optimality criteria based on the input/output-referred signal-to-noise ratio (SNR) of the independently measured pixels can lead to flawed decisions because it cannot encapsulate the location of the saturated pixels. We overcome the issue by proposing a new concept known as the spatially varying exposure risk (SVE-Risk) which is a pseudo-idealistic quantification of the amount of recoverable pixels. We present an efficient enumeration algorithm to select the optimal multiplex patterns. (ii) We report a design universality observation that the design of the multiplex pattern can be decoupled from the image reconstruction algorithm. This is a significant departure from the recent literature that the multiplex pattern should be jointly optimized with the reconstruction algorithm. Our finding suggests that in the context of exposure multiplexing, an end-to-end training may not be necessary.

In this paper, we study how pretraining label granularity affects the generalization of deep neural networks in image classification tasks. We focus on the "fine-to-coarse" transfer learning setting where the pretraining label is more fine-grained than that of the target problem. We experiment with this method using the label hierarchy of iNaturalist 2021, and observe a 8.76% relative improvement of the error rate over the baseline. We find the following conditions are key for the improvement: 1) the pretraining dataset has a strong and meaningful label hierarchy, 2) its label function strongly aligns with that of the target task, and most importantly, 3) an appropriate level of pretraining label granularity is chosen. The importance of pretraining label granularity is further corroborated by our transfer learning experiments on ImageNet. Most notably, we show that pretraining at the leaf labels of ImageNet21k produces better transfer results on ImageNet1k than pretraining at other coarser granularity levels, which supports the common practice. Theoretically, through an analysis on a two-layer convolutional ReLU network, we prove that: 1) models trained on coarse-grained labels only respond strongly to the common or "easy-to-learn" features; 2) with the dataset satisfying the right conditions, fine-grained pretraining encourages the model to also learn rarer or "harder-to-learn" features well, thus improving the model's generalization.

While today's high dynamic range (HDR) image fusion algorithms are capable of blending multiple exposures, the acquisition is often controlled so that the dynamic range within one exposure is narrow. For HDR imaging in photon-limited situations, the dynamic range can be enormous and the noise within one exposure is spatially varying. Existing image denoising algorithms and HDR fusion algorithms both fail to handle this situation, leading to severe limitations in low-light HDR imaging. This paper presents two contributions. Firstly, we identify the source of the problem. We find that the issue is associated with the co-existence of (1) spatially varying signal-to-noise ratio, especially the excessive noise due to very dark regions, and (2) a wide luminance range within each exposure. We show that while the issue can be handled by a bank of denoisers, the complexity is high. Secondly, we propose a new method called the spatially varying high dynamic range (SV-HDR) fusion network to simultaneously denoise and fuse images. We introduce a new exposure-shared block within our custom-designed multi-scale transformer framework. In a variety of testing conditions, the performance of the proposed SV-HDR is better than the existing methods.

Image restoration algorithms for atmospheric turbulence are known to be much more challenging to design than traditional ones such as blur or noise because the distortion caused by the turbulence is an entanglement of spatially varying blur, geometric distortion, and sensor noise. Existing CNN-based restoration methods built upon convolutional kernels with static weights are insufficient to handle the spatially dynamical atmospheric turbulence effect. To address this problem, in this paper, we propose a physics-inspired transformer model for imaging through atmospheric turbulence. The proposed network utilizes the power of transformer blocks to jointly extract a dynamical turbulence distortion map and restore a turbulence-free image. In addition, recognizing the lack of a comprehensive dataset, we collect and present two new real-world turbulence datasets that allow for evaluation with both classical objective metrics (e.g., PSNR and SSIM) and a new task-driven metric using text recognition accuracy. Both real testing sets and all related code will be made publicly available., This paper is accepted as a poster at ECCV 2022

Restoring images distorted by atmospheric turbulence is a long-standing problem due to the spatially varying nature of the distortion, nonlinearity of the image formation process, and scarcity of training and testing data. Existing methods often have strong statistical assumptions on the distortion model which in many cases will lead to a limited performance in real-world scenarios as they do not generalize. To overcome the challenge, this paper presents an end-to-end physics-driven approach that is efficient and can generalize to real-world turbulence. On the data synthesis front, we significantly increase the image resolution that can be handled by the SOTA turbulence simulator by approximating the random field via wide-sense stationarity. The new data synthesis process enables the generation of large-scale multi-level turbulence and ground truth pairs for training. On the network design front, we propose the turbulence mitigation transformer (TMT), a two stage U-Net shaped multi-frame restoration network which has a noval efficient self-attention mechanism named temporal channel joint attention (TCJA). We also introduce a new training scheme that is enabled by the new simulator, and we design new transformer units to reduce the memory consumption. Experimental results on both static and dynamic scenes are promising, including various real turbulence scenarios., Comment: 13 pages, 12 figures, project page: https://xg416.github.io/TMT/

Fast and accurate simulation of imaging through atmospheric turbulence is essential for developing turbulence mitigation algorithms. Recognizing the limitations of previous approaches, we introduce a new concept known as the phase-to-space (P2S) transform to significantly speed up the simulation. P2S is build upon three ideas: (1) reformulating the spatially varying convolution as a set of invariant convolutions with basis functions, (2) learning the basis function via the known turbulence statistics models, (3) implementing the P2S transform via a light-weight network that directly convert the phase representation to spatial representation. The new simulator offers 300x -- 1000x speed up compared to the mainstream split-step simulators while preserving the essential turbulence statistics., The paper will be published at the ICCV 2021

Identification of high-risk driving situations is generally approached through collision risk estimation or accident pattern recognition. In this work, we approach the problem from the perspective of subjective risk. We operationalize subjective risk assessment by predicting driver behavior changes and identifying the cause of changes. To this end, we introduce a new task called driver-centric risk object identification (DROID), which uses egocentric video to identify object(s) influencing a driver's behavior, given only the driver's response as the supervision signal. We formulate the task as a cause-effect problem and present a novel two-stage DROID framework, taking inspiration from models of situation awareness and causal inference. A subset of data constructed from the Honda Research Institute Driving Dataset (HDD) is used to evaluate DROID. We demonstrate state-of-the-art DROID performance, even compared with strong baseline models using this dataset. Additionally, we conduct extensive ablative studies to justify our design choices. Moreover, we demonstrate the applicability of DROID for risk assessment., Comment: Submitted to TPAMI

When training an estimator such as a neural network for tasks like image denoising, it is often preferred to train one estimator and apply it to all noise levels. The de facto training protocol to achieve this goal is to train the estimator with noisy samples whose noise levels are uniformly distributed across the range of interest. However, why should we allocate the samples uniformly? Can we have more training samples that are less noisy, and fewer samples that are more noisy? What is the optimal distribution? How do we obtain such a distribution? The goal of this paper is to address this training sample distribution problem from a minimax risk optimization perspective. We derive a dual ascent algorithm to determine the optimal sampling distribution of which the convergence is guaranteed as long as the set of admissible estimators is closed and convex. For estimators with non-convex admissible sets such as deep neural networks, our dual formulation converges to a solution of the convex relaxation. We discuss how the algorithm can be implemented in practice. We evaluate the algorithm on linear estimators and deep networks., Published in the 37th International Conference on Machine Learning (ICML) 2020

Imaging in low light is difficult because the number of photons arriving at the sensor is low. Imaging dynamic scenes in low-light environments is even more difficult because as the scene moves, pixels in adjacent frames need to be aligned before they can be denoised. Conventional CMOS image sensors (CIS) are at a particular disadvantage in dynamic low-light settings because the exposure cannot be too short lest the read noise overwhelms the signal. We propose a solution using Quanta Image Sensors (QIS) and present a new image reconstruction algorithm. QIS are single-photon image sensors with photon counting capabilities. Studies over the past decade have confirmed the effectiveness of QIS for low-light imaging but reconstruction algorithms for dynamic scenes in low light remain an open problem. We fill the gap by proposing a student-teacher training protocol that transfers knowledge from a motion teacher and a denoising teacher to a student network. We show that dynamic scenes can be reconstructed from a burst of frames at a photon level of 1 photon per pixel per frame. Experimental results confirm the advantages of the proposed method compared to existing methods., Comment: Published in the 16th European Conference on Computer Vision (ECCV) 2020

State-of-the-art image classifiers are trained and tested using well-illuminated images. These images are typically captured by CMOS image sensors with at least tens of photons per pixel. However, in dark environments when the photon flux is low, image classification becomes difficult because the measured signal is suppressed by noise. In this paper, we present a new low-light image classification solution using Quanta Image Sensors (QIS). QIS are a new type of image sensors that possess photon counting ability without compromising on pixel size and spatial resolution. Numerous studies over the past decade have demonstrated the feasibility of QIS for low-light imaging, but their usage for image classification has not been studied. This paper fills the gap by presenting a student-teacher learning scheme which allows us to classify the noisy QIS raw data. We show that with student-teacher learning, we are able to achieve image classification at a photon level of one photon per pixel or lower. Experimental results verify the effectiveness of the proposed method compared to existing solutions., Comment: Published in the 16th European Conference on Computer Vision (ECCV) 2020

State-of-the-art atmospheric turbulence image restoration methods utilize standard image processing tools such as optical flow, lucky region and blind deconvolution to restore the images. While promising results have been reported over the past decade, many of the methods are agnostic to the physical model that generates the distortion. In this paper, we revisit the turbulence restoration problem by analyzing the reference frame generation and the blind deconvolution steps in a typical restoration pipeline. By leveraging tools in large deviation theory, we rigorously prove the minimum number of frames required to generate a reliable reference for both static and dynamic scenes. We discuss how a turbulence agnostic model can lead to potential flaws, and how to configure a simple spatial-temporal non-local weighted averaging method to generate references. For blind deconvolution, we present a new data-driven prior by analyzing the distributions of the point spread functions. We demonstrate how a simple prior can outperform state-of-the-art blind deconvolution methods.

Regularized inversion methods for image reconstruction are used widely due to their tractability and ability to combine complex physical sensor models with useful regularity criteria. Such methods motivated the recently developed Plug-and-Play prior method, which provides a framework to use advanced denoising algorithms as regularizers in inversion. However, the need to formulate regularized inversion as the solution to an optimization problem limits the possible regularity conditions and physical sensor models. In this paper, we introduce Consensus Equilibrium (CE), which generalizes regularized inversion to include a much wider variety of both forward components and prior components without the need for either to be expressed with a cost function. CE is based on the solution of a set of equilibrium equations that balance data fit and regularity. In this framework, the problem of MAP estimation in regularized inversion is replaced by the problem of solving these equilibrium equations, which can be approached in multiple ways. The key contribution of CE is to provide a novel framework for fusing multiple heterogeneous models of physical sensors or models learned from data. We describe the derivation of the CE equations and prove that the solution of the CE equations generalizes the standard MAP estimate under appropriate circumstances. We also discuss algorithms for solving the CE equations, including ADMM with a novel form of preconditioning and Newton's method. We give examples to illustrate consensus equilibrium and the convergence properties of these algorithms and demonstrate this method on some toy problems and on a denoising example in which we use an array of convolutional neural network denoisers, none of which is tuned to match the noise level in a noisy image but which in consensus can achieve a better result than any of them individually., 17 pages

Alternating direction method of multiplier (ADMM) is a widely used algorithm for solving constrained optimization problems in image restoration. Among many useful features, one critical feature of the ADMM algorithm is its modular structure which allows one to plug in any off-the-shelf image denoising algorithm for a subproblem in the ADMM algorithm. Because of the plug-in nature, this type of ADMM algorithms is coined the name "Plug-and-Play ADMM". Plug-and-Play ADMM has demonstrated promising empirical results in a number of recent papers. However, it is unclear under what conditions and by using what denoising algorithms would it guarantee convergence. Also, since Plug-and-Play ADMM uses a specific way to split the variables, it is unclear if fast implementation can be made for common Gaussian and Poissonian image restoration problems. In this paper, we propose a Plug-and-Play ADMM algorithm with provable fixed point convergence. We show that for any denoising algorithm satisfying an asymptotic criteria, called bounded denoisers, Plug-and-Play ADMM converges to a fixed point under a continuation scheme. We also present fast implementations for two image restoration problems on super-resolution and single-photon imaging. We compare Plug-and-Play ADMM with state-of-the-art algorithms in each problem type, and demonstrate promising experimental results of the algorithm., 14 pages

This paper studies a type of image priors that are constructed implicitly through the alternating direction method of multiplier (ADMM) algorithm, called the algorithm-induced prior. Different from classical image priors which are defined before running the reconstruction algorithm, algorithm-induced priors are defined by the denoising procedure used to replace one of the two modules in the ADMM algorithm. Since such prior is not explicitly defined, analyzing the performance has been difficult in the past. Focusing on the class of symmetric smoothing filters, this paper presents an explicit expression of the prior induced by the ADMM algorithm. The new prior is reminiscent to the conventional graph Laplacian but with stronger reconstruction performance. It can also be shown that the overall reconstruction has an efficient closed-form implementation if the associated symmetric smoothing filter is low rank. The results are validated with experiments on image inpainting.

Non-parametric approaches for analyzing network data based on exchangeable graph models (ExGM) have recently gained interest. The key object that defines an ExGM is often referred to as a graphon. This non-parametric perspective on network modeling poses challenging questions on how to make inference on the graphon underlying observed network data. In this paper, we propose a computationally efficient procedure to estimate a graphon from a set of observed networks generated from it. This procedure is based on a stochastic blockmodel approximation (SBA) of the graphon. We show that, by approximating the graphon with a stochastic block model, the graphon can be consistently estimated, that is, the estimation error vanishes as the size of the graph approaches infinity., 20 pages, 4 figures, 2 algorithms. Neural Information Processing Systems (NIPS), 2013