Spectral and spatial correlation in hyperspectral images (HSIs) can be exploited in HSI processing because it directly induces a sparse and low-rank prior via linear transformations. Researchers have used the sparse and low-rank prior as an image prior for HSI restoration, such as denoising, deblurring, and super-resolution. This paper proposes a HSI denoising method that incorporates a sparse and low-rank prior with a deep image prior (DIP). The sparse and low-rank prior is obtained using the 2-dimensional discrete wavelet transform (2-D DWT), and singular value decomposition (SVD), while the DIP is provided by the structure of a convolutional neural network (CNN). The combination of a sparse and low-rank prior with a DIP views the CNN-based denoising method similar to a model-based method, inheriting the advantages of both model-based and CNN-based methods. Experimental results with simulated and real HSI datasets show that the proposed method outperforms the conventional sparse and low-rank based methods in both quantitative and qualitative performance. Codes are available at https://github.com/hvn2/DIP-SLR
The Sentinel-2 (S2) constellation provides multispectral images at 10 m, 20 m, and 60 m resolution bands. Obtaining all bands at 10 m resolution would benefit many applications. Recently, many model-based and deep learning (DL)-based sharpening methods have been proposed. However, the downside of those methods is that the DL-based methods need to be trained separately for the 20 m and the 60 m bands in a supervised manner at reduced resolution, while the model-based methods heavily depend on the hand-crafted image priors. To break the gap, this article proposes a novel unsupervised DL-based S2 sharpening method using a single convolutional neural network (CNN) to sharpen the 20 and 60 m bands at the same time at full resolution. The proposed method replaces the hand-crafted image prior by the deep image prior (DIP) provided by a CNN structure whose parameters are easily optimized using a DL optimizer. We also incorporate the modulation transfer function-based degradation model as a network layer, and add all bands to both network input and output. This setting improves the DIP and exploits the advantage of multitask learning since all S2 bands are highly correlated. Extensive experiments with real S2 data show that our proposed method outperforms competitive methods for reduced-resolution evaluation and yields very high quality sharpened image for full-resolution evaluation.
Mobile channel modeling has always been the core part for design, deployment and optimization of communication system, especially in 5G and beyond era. Deterministic channel modeling could precisely achieve mobile channel description, however with defects of equipment and time consuming. In this paper, we proposed a novel super resolution (SR) model for cluster characteristics prediction. The model is based on deep neural networks with residual connection. A series of simulations at 3.5 GHz are conducted by a three-dimensional ray tracing (RT) simulator in diverse scenarios. Cluster characteristics are extracted and corresponding data sets are constructed to train the model. Experiments demonstrate that the proposed SR approach could achieve better power and cluster location prediction performance than traditional interpolation method and the root mean square error (RMSE) drops by 51% and 78% relatively. Channel impulse response (CIR) is reconstructed based on cluster characteristics, which could match well with the multi-path component (MPC). The proposed method can be used to efficiently and accurately generate big data of mobile channel, which significantly reduces the computation time of RT-only.
This paper presents a stochastic differential equation (SDE) approach for general-purpose image restoration. The key construction consists in a mean-reverting SDE that transforms a high-quality image into a degraded counterpart as a mean state with fixed Gaussian noise. Then, by simulating the corresponding reverse-time SDE, we are able to restore the origin of the low-quality image without relying on any task-specific prior knowledge. Crucially, the proposed mean-reverting SDE has a closed-form solution, allowing us to compute the ground truth time-dependent score and learn it with a neural network. Moreover, we propose a maximum likelihood objective to learn an optimal reverse trajectory that stabilizes the training and improves the restoration results. The experiments show that our proposed method achieves highly competitive performance in quantitative comparisons on image deraining, deblurring, and denoising, setting a new state-of-the-art on two deraining datasets. Finally, the general applicability of our approach is further demonstrated via qualitative results on image super-resolution, inpainting, and dehazing. Code is available at https://github.com/Algolzw/image-restoration-sde.
Guided diffusion is a technique for conditioning the output of a diffusion model at sampling time without retraining the network for each specific task. One drawback of diffusion models, however, is their slow sampling process. Recent techniques can accelerate unguided sampling by applying high-order numerical methods to the sampling process when viewed as differential equations. On the contrary, we discover that the same techniques do not work for guided sampling, and little has been explored about its acceleration. This paper explores the culprit of this problem and provides a solution based on operator splitting methods, motivated by our key finding that classical high-order numerical methods are unsuitable for the conditional function. Our proposed method can re-utilize the high-order methods for guided sampling and can generate images with the same quality as a 250-step DDIM baseline using 32-58% less sampling time on ImageNet256. We also demonstrate usage on a wide variety of conditional generation tasks, such as text-to-image generation, colorization, inpainting, and super-resolution.
This paper surveys machine-learning-based super-resolution reconstruction for vortical flows. Super resolution aims to find the high-resolution flow fields from low-resolution data and is generally an approach used in image reconstruction. In addition to surveying a variety of recent super-resolution applications, we provide case studies of super-resolution analysis for an example of two-dimensional decaying isotropic turbulence. We demonstrate that physics-inspired model designs enable successful reconstruction of vortical flows from spatially limited measurements. We also discuss the challenges and outlooks of machine-learning-based super-resolution analysis for fluid flow applications. The insights gained from this study can be leveraged for super-resolution analysis of numerical and experimental flow data.
Multispectral Sentinel-2 images are a valuable source of Earth observation data, however spatial resolution of their spectral bands limited to 10 m, 20 m, and 60 m ground sampling distance remains insufficient in many cases. This problem can be addressed with super-resolution, aimed at reconstructing a high-resolution image from a low-resolution observation. For Sentinel-2, spectral information fusion allows for enhancing the 20 m and 60 m bands to the 10 m resolution. Also, there were attempts to combine multitemporal stacks of individual Sentinel-2 bands, however these two approaches have not been combined so far. In this paper, we introduce DeepSent -- a new deep network for super-resolving multitemporal series of multispectral Sentinel-2 images. It is underpinned with information fusion performed simultaneously in the spectral and temporal dimensions to generate an enlarged multispectral image. In our extensive experimental study, we demonstrate that our solution outperforms other state-of-the-art techniques that realize either multitemporal or multispectral data fusion. Furthermore, we show that the advantage of DeepSent results from how these two fusion types are combined in a single architecture, which is superior to performing such fusion in a sequential manner. Importantly, we have applied our method to super-resolve real-world Sentinel-2 images, enhancing the spatial resolution of all the spectral bands to 3.3 m nominal ground sampling distance, and we compare the outcome with very high-resolution WorldView-2 images. We will publish our implementation upon paper acceptance, and we expect it will increase the possibilities of exploiting super-resolved Sentinel-2 images in real-life applications.
Currently, applying diffusion models in pixel space of high resolution images is difficult. Instead, existing approaches focus on diffusion in lower dimensional spaces (latent diffusion), or have multiple super-resolution levels of generation referred to as cascades. The downside is that these approaches add additional complexity to the diffusion framework. This paper aims to improve denoising diffusion for high resolution images while keeping the model as simple as possible. The paper is centered around the research question: How can one train a standard denoising diffusion models on high resolution images, and still obtain performance comparable to these alternate approaches? The four main findings are: 1) the noise schedule should be adjusted for high resolution images, 2) It is sufficient to scale only a particular part of the architecture, 3) dropout should be added at specific locations in the architecture, and 4) downsampling is an effective strategy to avoid high resolution feature maps. Combining these simple yet effective techniques, we achieve state-of-the-art on image generation among diffusion models without sampling modifiers on ImageNet.
In recent years, limited research has discussed the loss function in the super-resolution process. The majority of those studies have only used perceptual similarity conventionally. This is while the development of appropriate loss can improve the quality of other methods as well. In this article, a new weighting method for pixel-wise loss is proposed. With the help of this method, it is possible to use trainable weights based on the general structure of the image and its perceptual features while maintaining the advantages of pixel-wise loss. Also, a criterion for comparing weights of loss is introduced so that the weights can be estimated directly by a convolutional neural network. In addition, in this article, the expectation-maximization method is used for the simultaneous estimation super-resolution network and weighting network. In addition, a new activation function, called "FixedSum", is introduced which can keep the sum of all components of vector constants while keeping the output components between zero and one. As experimental results shows, weighted loss by the proposed method leads to better results than the unweighted loss and weighted loss based on uncertainty in both signal-to-noise and perceptual similarity senses on the state-of-the-art networks. Code is available online.
Closed-loop negative feedback mechanism is extensively utilized in automatic control systems and brings about extraordinary dynamic and static performance. In order to further improve the reconstruction capability of current methods of compressed image super-resolution, a circular Swin2SR (CSwin2SR) approach is proposed. The CSwin2SR contains a serial Swin2SR for initial super-resolution reestablishment and circular Swin2SR for enhanced super-resolution reestablishment. Simulated experimental results show that the proposed CSwin2SR dramatically outperforms the classical Swin2SR in the capacity of super-resolution recovery. On DIV2K test and valid datasets, the average increment of PSNR is greater than 0.18 dB and the related average increment of SSIM is greater than 0.01.
The quality of the video stream is key to neural network-based video analytics. However, low-quality video is inevitably collected by existing surveillance systems because of poor quality cameras or over-compressed/pruned video streaming protocols, e.g., as a result of upstream bandwidth limit. To address this issue, existing studies use quality enhancers (e.g., neural super-resolution) to improve the quality of videos (e.g., resolution) and eventually ensure inference accuracy. Nevertheless, directly applying quality enhancers does not work in practice because it will introduce unacceptable latency. In this paper, we present AccDecoder, a novel accelerated decoder for real-time and neural-enhanced video analytics. AccDecoder can select a few frames adaptively via Deep Reinforcement Learning (DRL) to enhance the quality by neural super-resolution and then up-scale the unselected frames that reference them, which leads to 6-21% accuracy improvement. AccDecoder provides efficient inference capability via filtering important frames using DRL for DNN-based inference and reusing the results for the other frames via extracting the reference relationship among frames and blocks, which results in a latency reduction of 20-80% than baselines.
This paper proposes a super-resolution harmonic retrieval method for uncorrelated strictly non-circular signals, whose covariance and pseudo-covariance present Toeplitz and Hankel structures, respectively. Accordingly, the augmented covariance matrix constructed by the covariance and pseudo-covariance matrices is not only low rank but also jointly Toeplitz-Hankel structured. To efficiently exploit such a desired structure for high estimation accuracy, we develop a low-rank Toeplitz-Hankel covariance reconstruction (LRTHCR) solution employed over the augmented covariance matrix. Further, we design a fitting error constraint to flexibly implement the LRTHCR algorithm without knowing the noise statistics. In addition, performance analysis is provided for the proposed LRTHCR in practical settings. Simulation results reveal that the LRTHCR outperforms the benchmark methods in terms of lower estimation errors.
In this paper, we study the problem of efficiently and effectively embedding the high-dimensional spatio-spectral information of hyperspectral (HS) images, guided by feature diversity. Specifically, based on the theoretical formulation that feature diversity is correlated with the rank of the unfolded kernel matrix, we rectify 3D convolution by modifying its topology to enhance the rank upper-bound. This modification yields a rank-enhanced spatial-spectral symmetrical convolution set (ReS$^3$-ConvSet), which not only learns diverse and powerful feature representations but also saves network parameters. Additionally, we also propose a novel diversity-aware regularization (DA-Reg) term that directly acts on the feature maps to maximize independence among elements. To demonstrate the superiority of the proposed ReS$^3$-ConvSet and DA-Reg, we apply them to various HS image processing and analysis tasks, including denoising, spatial super-resolution, and classification. Extensive experiments show that the proposed approaches outperform state-of-the-art methods both quantitatively and qualitatively to a significant extent. The code is publicly available at https://github.com/jinnh/ReSSS-ConvSet.
Implicit neural representations (INRs) have recently advanced numerous vision-related areas. INR performance depends strongly on the choice of the nonlinear activation function employed in its multilayer perceptron (MLP) network. A wide range of nonlinearities have been explored, but, unfortunately, current INRs designed to have high accuracy also suffer from poor robustness (to signal noise, parameter variation, etc.). Inspired by harmonic analysis, we develop a new, highly accurate and robust INR that does not exhibit this tradeoff. Wavelet Implicit neural REpresentation (WIRE) uses a continuous complex Gabor wavelet activation function that is well-known to be optimally concentrated in space-frequency and to have excellent biases for representing images. A wide range of experiments (image denoising, image inpainting, super-resolution, computed tomography reconstruction, image overfitting, and novel view synthesis with neural radiance fields) demonstrate that WIRE defines the new state of the art in INR accuracy, training time, and robustness.
As the most common representation for 3D shapes, mesh is often stored discretely with arrays of vertices and faces. However, 3D shapes in the real world are presented continuously. In this paper, we propose to learn a continuous representation for meshes with fixed topology, a common and practical setting in many faces-, hand-, and body-related applications. First, we split the template into multiple closed manifold genus-0 meshes so that each genus-0 mesh can be parameterized onto the unit sphere. Then we learn spherical implicit surface (SIS), which takes a spherical coordinate and a global feature or a set of local features around the coordinate as inputs, predicting the vertex corresponding to the coordinate as an output. Since the spherical coordinates are continuous, SIS can depict a mesh in an arbitrary resolution. SIS representation builds a bridge between discrete and continuous representation in 3D shapes. Specifically, we train SIS networks in a self-supervised manner for two tasks: a reconstruction task and a super-resolution task. Experiments show that our SIS representation is comparable with state-of-the-art methods that are specifically designed for meshes with a fixed resolution and significantly outperforms methods that work in arbitrary resolutions.
While convolutional neural networks (CNNs) demonstrate outstanding performance on computer vision tasks, their computational costs remain high. Several techniques are used to reduce these costs, like reducing channel count, and using separable and depthwise separable convolutions. This paper reduces computational costs by introducing a novel architecture, axial CNNs, which replaces spatial 2D convolution operations with two consecutive depthwise separable 1D operations. The axial CNNs are predicated on the assumption that the dataset supports approximately separable convolution operations with little or no loss of training accuracy. Deep axial separable CNNs still suffer from gradient problems when training deep networks. We modify the construction of axial separable CNNs with residual connections to improve the performance of deep axial architectures and introduce our final novel architecture namely residual axial networks (RANs). Extensive benchmark evaluation shows that RANs achieve at least 1% higher performance with about 77%, 86%, 75%, and 34% fewer parameters and about 75%, 80%, 67%, and 26% fewer flops than ResNets, wide ResNets, MobileNets, and SqueezeNexts on CIFAR benchmarks, SVHN, and Tiny ImageNet image classification datasets. Moreover, our proposed RANs improve deep recursive residual networks performance with 94% fewer parameters on the image super-resolution dataset.
As an emerging approach, deep learning plays an increasingly influential role in channel modeling. Traditional ray tracing (RT) methods of channel modeling tend to be inefficient and expensive. In this paper, we present a super-resolution (SR) model for channel characteristics. Residual connection and attention mechanism are applied to this convolutional neural network (CNN) model. Experiments prove that the proposed model can reduce the noise interference generated in the SR process and solve the problem of low efficiency of RT. The mean absolute error of our channel SR model on the PL achieves the effect of 2.82 dB with scale factor 2, the same accuracy as RT took only 52\% of the time in theory. Compared with vision transformer (ViT), the proposed model also demonstrates less running time and computing cost in SR of channel characteristics.
The application of deep learning techniques has greatly enhanced holographic imaging capabilities, leading to improved phase recovery and image reconstruction. Here, we introduce a deep neural network termed enhanced Fourier Imager Network (eFIN) as a highly generalizable framework for hologram reconstruction with pixel super-resolution and image autofocusing. Through holographic microscopy experiments involving lung, prostate and salivary gland tissue sections and Papanicolau (Pap) smears, we demonstrate that eFIN has a superior image reconstruction quality and exhibits external generalization to new types of samples never seen during the training phase. This network achieves a wide autofocusing axial range of 0.35 mm, with the capability to accurately predict the hologram axial distances by physics-informed learning. eFIN enables 3x pixel super-resolution imaging and increases the space-bandwidth product of the reconstructed images by 9-fold with almost no performance loss, which allows for significant time savings in holographic imaging and data processing steps. Our results showcase the advancements of eFIN in pushing the boundaries of holographic imaging for various applications in e.g., quantitative phase imaging and label-free microscopy.
We propose an effective lightweight dynamic local and global self-attention network (DLGSANet) to solve image super-resolution. Our method explores the properties of Transformers while having low computational costs. Motivated by the network designs of Transformers, we develop a simple yet effective multi-head dynamic local self-attention (MHDLSA) module to extract local features efficiently. In addition, we note that existing Transformers usually explore all similarities of the tokens between the queries and keys for the feature aggregation. However, not all the tokens from the queries are relevant to those in keys, using all the similarities does not effectively facilitate the high-resolution image reconstruction. To overcome this problem, we develop a sparse global self-attention (SparseGSA) module to select the most useful similarity values so that the most useful global features can be better utilized for the high-resolution image reconstruction. We develop a hybrid dynamic-Transformer block(HDTB) that integrates the MHDLSA and SparseGSA for both local and global feature exploration. To ease the network training, we formulate the HDTBs into a residual hybrid dynamic-Transformer group (RHDTG). By embedding the RHDTGs into an end-to-end trainable network, we show that our proposed method has fewer network parameters and lower computational costs while achieving competitive performance against state-of-the-art ones in terms of accuracy. More information is available at https://neonleexiang.github.io/DLGSANet/
Sparse arrays have emerged as a popular alternative to the conventional uniform linear array (ULA) due to the enhanced degrees of freedom (DOF) and superior resolution offered by them. In the passive setting, these advantages are realized by leveraging correlation between the received signals at different sensors. This has led to the belief that sparse arrays require a large number of temporal measurements to reliably estimate parameters of interest from these correlations, and therefore they may not be preferred in the sample-starved regime. In this paper, we debunk this myth by performing a rigorous non-asymptotic analysis of the Coarray ESPRIT algorithm. This seemingly counter-intuitive result is a consequence of the scaling of the singular value of the coarray manifold, which compensates for the potentially large covariance estimation error in the limited snapshot regime. Specifically, we show that for a nested array operating in the regime of fewer sources than sensors ($S=O(1)$), it is possible to bound the matching distance error between the estimated and true directions of arrival (DOAs) by an arbitrarily small quantity ($\epsilon$) with high probability, provided (i) the number of temporal snapshots ($L$) scales only logarithmically with the number of sensors ($P$), i.e. $L=\Omega(\ln(P)/\epsilon^2)$, and (ii) a suitable separation condition is satisfied. Our results also formally prove the well-known empirical resolution benefits of sparse arrays, by establishing that the minimum separation between sources can be $\Omega(1/P^2)$, as opposed to separation $\Omega(1/P)$ required by a ULA with the same number of sensors. Our sample complexity expression reveals the dependence on other key model parameters such as SNR and the dynamic range of the source powers. This enables us to establish the superior noise-resilience of nested arrays both theoretically and empirically.