A novel approach towards depth map super-resolution using multi-view uncalibrated photometric stereo is presented. Practically, an LED light source is attached to a commodity RGB-D sensor and is used to capture objects from multiple viewpoints with unknown motion. This non-static camera-to-object setup is described with a nonconvex variational approach such that no calibration on lighting or camera motion is required due to the formulation of an end-to-end joint optimization problem. Solving the proposed variational model results in high resolution depth, reflectance and camera pose estimates, as we show on challenging synthetic and real-world datasets.
This paper presents a generative adversarial network based super-resolution (SR) approach (which is called as S2GAN) to enhance the spatial resolution of Sentinel-2 spectral bands. The proposed approach consists of two main steps. The first step aims to increase the spatial resolution of the bands with 20m and 60m spatial resolutions by the scaling factors of 2 and 6, respectively. To this end, we introduce a generator network that performs SR on the lower resolution bands with the guidance of the bands associated to 10m spatial resolution by utilizing the convolutional layers with residual connections and a long skip-connection between inputs and outputs. The second step aims to distinguish SR bands from their ground truth bands. This is achieved by the proposed discriminator network, which alternately characterizes the high level features of the two sets of bands and applying binary classification on the extracted features. Then, we formulate the adversarial learning of the generator and discriminator networks as a min-max game. In this learning procedure, the generator aims to produce realistic SR bands as much as possible so that the discriminator incorrectly classifies SR bands. Experimental results obtained on different Sentinel-2 images show the effectiveness of the proposed approach compared to both conventional and deep learning based SR approaches.
Computational imaging systems jointly design computation and hardware to retrieve information which is not traditionally accessible with standard imaging systems. Recently, critical aspects such as experimental design and image priors are optimized through deep neural networks formed by the unrolled iterations of classical physics-based reconstructions (termed physics-based networks). However, for real-world large-scale systems, computing gradients via backpropagation restricts learning due to memory limitations of graphical processing units. In this work, we propose a memory-efficient learning procedure that exploits the reversibility of the network's layers to enable data-driven design for large-scale computational imaging. We demonstrate our methods practicality on two large-scale systems: super-resolution optical microscopy and multi-channel magnetic resonance imaging.
Video-to-video translation is more difficult than image-to-image translation due to the temporal consistency problem that, if unaddressed, leads to distracting flickering effects. Although video models designed from scratch produce temporally consistent results, training them to match the vast visual knowledge captured by image models requires an intractable number of videos. To combine the benefits of image and video models, we propose an image-to-video model transfer method called Hyperconsistency (HyperCon) that transforms any well-trained image model into a temporally consistent video model without fine-tuning. HyperCon works by translating a temporally interpolated video frame-wise and then aggregating over temporally localized windows on the interpolated video. It handles both masked and unmasked inputs, enabling support for even more video-to-video translation tasks than prior image-to-video model transfer techniques. We demonstrate HyperCon on video style transfer and inpainting, where it performs favorably compared to prior state-of-the-art methods without training on a single stylized or incomplete video. Our project website is available at https://ryanszeto.com/projects/hypercon .
Recently, there is a vast interest in developing methods which are independent of the training samples such as deep image prior, zero-shot learning, and internal learning. The methods above are based on the common goal of maximizing image features learning from a single image despite inherent technical diversity. In this work, we bridge the gap between the various unsupervised approaches above and propose a general framework for image restoration and image retargeting. We use contextual feature learning and internal learning to improvise the structure similarity between the source and the target images. We perform image resize application in the following setups: classical image resize using super-resolution, a challenging image resize where the low-resolution image contains noise, and content-aware image resize using image retargeting. We also provide comparisons to the relevant state-of-the-art methods.
Recently, Convolutional Neural Networks (CNNs) have been successfully adopted to solve the ill-posed single image super-resolution (SISR) problem. A commonly used strategy to boost the performance of CNN-based SISR models is deploying very deep networks, which inevitably incurs many obvious drawbacks (e.g., a large number of network parameters, heavy computational loads, and difficult model training). In this paper, we aim to build more accurate and faster SISR models via developing better-performing feature extraction and fusion techniques. Firstly, we proposed a novel Orientation-Aware feature extraction and fusion Module (OAM), which contains a mixture of 1D and 2D convolutional kernels (i.e., 5 x 1, 1 x 5, and 3 x 3) for extracting orientation-aware features. Secondly, we adopt the channel attention mechanism as an effective technique to adaptively fuse features extracted in different directions and in hierarchically stacked convolutional stages. Based on these two important improvements, we present a compact but powerful CNN-based model for high-quality SISR via Channel Attention-based fusion of Orientation-Aware features (SISR-CA-OA). Extensive experimental results verify the superiority of the proposed SISR-CA-OA model, performing favorably against the state-of-the-art SISR models in terms of both restoration accuracy and computational efficiency. The source codes will be made publicly available.
Convolutional Neural Networks have reached extremely high performances on the Face Recognition task. Largely used datasets, such as VGGFace2, focus on gender, pose and age variations trying to balance them to achieve better results. However, the fact that images have different resolutions is not usually discussed and resize to 256 pixels before cropping is used. While specific datasets for very low resolution faces have been proposed, less attention has been payed on the task of cross-resolution matching. Such scenarios are of particular interest for forensic and surveillance systems in which it usually happens that a low-resolution probe has to be matched with higher-resolution galleries. While it is always possible to either increase the resolution of the probe image or to reduce the size of the gallery images, to the best of our knowledge an extensive experimentation of cross-resolution matching was missing in the recent deep learning based literature. In the context of low- and cross-resolution Face Recognition, the contributions of our work are: i) we proposed a training method to fine-tune a state-of-the-art model in order to make it able to extract resolution-robust deep features; ii) we tested our models on the benchmark datasets IJB-B/C considering images at both full and low resolutions in order to show the effectiveness of the proposed training algorithm. To the best of our knowledge, this is the first work testing extensively the performance of a FR model in a cross-resolution scenario; iii) we tested our models on the low resolution and low quality datasets QMUL-SurvFace and TinyFace and showed their superior performances, even though we did not train our model on low-resolution faces only and our main focus was cross-resolution; iv) we showed that our approach can be more effective with respect to preprocessing faces with super resolution techniques.
In speech synthesis and speech enhancement systems, melspectrograms need to be precise in acoustic representations. However, the generated spectrograms are over-smooth, that could not produce high quality synthesized speech. Inspired by image-to-image translation, we address this problem by using a learning-based post filter combining Pix2PixHD and ResUnet to reconstruct the mel-spectrograms together with super-resolution. From the resulting super-resolution spectrogram networks, we can generate enhanced spectrograms to produce high quality synthesized speech. Our proposed model achieves improved mean opinion scores (MOS) of 3.71 and 4.01 over baseline results of 3.29 and 3.84, while using vocoder Griffin-Lim and WaveNet, respectively.
Single image super resolution (SR) has seen major performance leaps in recent years. However, existing methods do not allow exploring the infinitely many plausible reconstructions that might have given rise to the observed low-resolution (LR) image. These different explanations to the LR image may dramatically vary in their textures and fine details, and may often encode completely different semantic information. In this paper, we introduce the task of explorable super resolution. We propose a framework comprising a graphical user interface with a neural network backend, allowing editing the SR output so as to explore the abundance of plausible HR explanations to the LR input. At the heart of our method is a novel module that can wrap any existing SR network, analytically guaranteeing that its SR outputs would precisely match the LR input, when downsampled. Besides its importance in our setting, this module is guaranteed to decrease the reconstruction error of any SR network it wraps, and can be used to cope with blur kernels that are different from the one the network was trained for. We illustrate our approach in a variety of use cases, ranging from medical imaging and forensics, to graphics.
Accurate and realistic simulation of high-dimensional medical images has become an important research area relevant to many AI-enabled healthcare applications. However, current state-of-the-art approaches lack the ability to produce satisfactory high-resolution and accurate subject-specific images. In this work, we present a deep learning framework, namely 4D-Degenerative Adversarial NeuroImage Net (4D-DANI-Net), to generate high-resolution, longitudinal MRI scans that mimic subject-specific neurodegeneration in ageing and dementia. 4D-DANI-Net is a modular framework based on adversarial training and a set of novel spatiotemporal, biologically-informed constraints. To ensure efficient training and overcome memory limitations affecting such high-dimensional problems, we rely on three key technological advances: i) a new 3D training consistency mechanism called Profile Weight Functions (PWFs), ii) a 3D super-resolution module and iii) a transfer learning strategy to fine-tune the system for a given individual. To evaluate our approach, we trained the framework on 9852 T1-weighted MRI scans from 876 participants in the Alzheimer's Disease Neuroimaging Initiative dataset and held out a separate test set of 1283 MRI scans from 170 participants for quantitative and qualitative assessment of the personalised time series of synthetic images. We performed three evaluations: i) image quality assessment; ii) quantifying the accuracy of regional brain volumes over and above benchmark models; and iii) quantifying visual perception of the synthetic images by medical experts. Overall, both quantitative and qualitative results show that 4D-DANI-Net produces realistic, low-artefact, personalised time series of synthetic T1 MRI that outperforms benchmark models.
An event camera detects per-pixel intensity difference and produces asynchronous event stream with low latency, high dynamic range, and low power consumption. As a trade-off, the event camera has low spatial resolution. We propose an end-to-end network to reconstruct high resolution, high dynamic range (HDR) images directly from the event stream. We evaluate our algorithm on both simulated and real-world sequences and verify that it captures fine details of a scene and outperforms the combination of the state-of-the-art event to image algorithms with the state-of-the-art super resolution schemes in many quantitative measures by large margins. We further extend our method by using the active sensor pixel (APS) frames or reconstructing images iteratively.
We aim to super-resolve digital paintings, synthesizing realistic details from high-resolution reference painting materials for very large scaling factors (e.g., 8X, 16X). However, previous single image super-resolution (SISR) methods would either lose textural details or introduce unpleasing artifacts. On the other hand, reference-based SR (Ref-SR) methods can transfer textures to some extent, but is still impractical to handle very large factors and keep fidelity with original input. To solve these problems, we propose an efficient high-resolution hallucination network for very large scaling factors with an efficient network structure and feature transferring. To transfer more detailed textures, we design a wavelet texture loss, which helps to enhance more high-frequency components. At the same time, to reduce the smoothing effect brought by the image reconstruction loss, we further relax the reconstruction constraint with a degradation loss which ensures the consistency between downscaled super-resolution results and low-resolution inputs. We also collected a high-resolution (e.g., 4K resolution) painting dataset PaintHD by considering both physical size and image resolution. We demonstrate the effectiveness of our method with extensive experiments on PaintHD by comparing with SISR and Ref-SR state-of-the-art methods.
The single image super-resolution task is one of the most examined inverse problems in the past decade. In the recent years, Deep Neural Networks (DNNs) have shown superior performance over alternative methods when the acquisition process uses a fixed known downsampling kernel-typically a bicubic kernel. However, several recent works have shown that in practical scenarios, where the test data mismatch the training data (e.g. when the downsampling kernel is not the bicubic kernel or is not available at training), the leading DNN methods suffer from a huge performance drop. Inspired by the literature on generalized sampling, in this work we propose a method for improving the performance of DNNs that have been trained with a fixed kernel on observations acquired by other kernels. For a known kernel, we design a closed-form correction filter that modifies the low-resolution image to match one which is obtained by another kernel (e.g. bicubic), and thus improves the results of existing pre-trained DNNs. For an unknown kernel, we extend this idea and propose an algorithm for blind estimation of the required correction filter. We show that our approach outperforms other super-resolution methods, which are designed for general downsampling kernels.
Normalizing Flows (NFs) are able to model complicated distributions p(y) with strong inter-dimensional correlations and high multimodality by transforming a simple base density p(z) through an invertible neural network under the change of variables formula. Such behavior is desirable in multivariate structured prediction tasks, where handcrafted per-pixel loss-based methods inadequately capture strong correlations between output dimensions. We present a study of conditional normalizing flows (CNFs), a class of NFs where the base density to output space mapping is conditioned on an input x, to model conditional densities p(y|x). CNFs are efficient in sampling and inference, they can be trained with a likelihood-based objective, and CNFs, being generative flows, do not suffer from mode collapse or training instabilities. We provide an effective method to train continuous CNFs for binary problems and in particular, we apply these CNFs to super-resolution and vessel segmentation tasks demonstrating competitive performance on standard benchmark datasets in terms of likelihood and conventional metrics.
Medical image synthesis has gained a great focus recently, especially after the introduction of Generative Adversarial Networks (GANs). GANs have been used widely to provide anatomically-plausible and diverse samples for augmentation and other applications, including segmentation and super resolution. In our previous work, Deep Convolutional GANs were used to generate synthetic mammogram lesions, masses mainly, that could enhance the classification performance in imbalanced datasets. In this new work, a deeper investigation was carried out to explore other aspects of the generated images evaluation, i.e., realism, feature space distribution, and observers studies. t-Stochastic Neighbor Embedding (t-SNE) was used to reduce the dimensionality of real and fake images to enable 2D visualisations. Additionally, two expert radiologists performed a realism-evaluation study. Visualisations showed that the generated images have a similar feature distribution of the real ones, avoiding outliers. Moreover, Receiver Operating Characteristic (ROC) curve showed that the radiologists could not, in many cases, distinguish between synthetic and real lesions, giving 48% and 61% accuracies in a balanced sample set.
Recent works propose neural network- (NN-) inspired analog-to-digital converters (NNADCs) and demonstrate their great potentials in many emerging applications. These NNADCs often rely on resistive random-access memory (RRAM) devices to realize the NN operations and require high-precision RRAM cells (6~12-bit) to achieve a moderate quantization resolution (4~8-bit). Such optimistic assumption of RRAM resolution, however, is not supported by fabrication data of RRAM arrays in large-scale production process. In this paper, we propose an NN-inspired super-resolution ADC based on low-precision RRAM devices by taking the advantage of a co-design methodology that combines a pipelined hardware architecture with a custom NN training framework. Results obtained from SPICE simulations demonstrate that our method leads to robust design of a 14-bit super-resolution ADC using 3-bit RRAM devices with improved power and speed performance and competitive figure-of-merits (FoMs). In addition to the linear uniform quantization, the proposed ADC can also support configurable high-resolution nonlinear quantization with high conversion speed and low conversion energy, enabling future intelligent analog-to-information interfaces for near-sensor analytics and processing.
Purpose: Probe-based Confocal Laser Endomicroscopy (pCLE) enables performing an optical biopsy, providing real-time microscopic images, via a probe. pCLE probes consist of multiple optical fibres arranged in a bundle, which taken together generate signals in an irregularly sampled pattern. Current pCLE reconstruction is based on interpolating irregular signals onto an over-sampled Cartesian grid, using a naive linear interpolation. It was shown that Convolutional Neural Networks (CNNs) could improve pCLE image quality. Although classical CNNs were applied to pCLE, input data were limited to reconstructed images in contrast to irregular data produced by pCLE. Methods: We compare pCLE reconstruction and super-resolution (SR) methods taking irregularly sampled or reconstructed pCLE images as input. We also propose to embed a Nadaraya-Watson (NW) kernel regression into the CNN framework as a novel trainable CNN layer. Using the NW layer and exemplar-based super-resolution, we design an NWNetSR architecture that allows for reconstructing high-quality pCLE images directly from the irregularly sampled input data. We created synthetic sparse pCLE images to evaluate our methodology. Results: The results were validated through an image quality assessment based on a combination of the following metrics: Peak signal-to-noise ratio, the Structural Similarity Index. Conclusion: Both dense and sparse CNNs outperform the reconstruction method currently used in the clinic. The main contributions of our study are a comparison of sparse and dense approach in pCLE image reconstruction, implementing trainable generalised NW kernel regression, and adaptation of synthetic data for training pCLE SR.
Synthesizing a densely sampled light field from a single image is highly beneficial for many applications. Moreover, jointly solving both angular and spatial super-resolution problem also introduces new possibilities in light field imaging. The conventional method relies on physical-based rendering and a secondary network to solve the angular super-resolution problem. In addition, pixel-based loss limits the network capability to infer scene geometry globally. In this paper, we show that both super-resolution problems can be solved jointly from a single image by proposing a single end-to-end deep neural network that does not require a physical-based approach. Two novel loss functions based on known light field domain knowledge are proposed to enable the network to preserve the spatio-angular consistency between sub-aperture images. Experimental results show that the proposed model successfully synthesizes dense high resolution light field and it outperforms the state-of-the-art method in both quantitative and qualitative criteria. The method can be generalized to arbitrary scenes, rather than focusing on a particular subject. The synthesized light field can be used for various applications, such as depth estimation and refocusing.
As a domain-specific super-resolution problem, facial image hallucination has enjoyed a series of breakthroughs thanks to the advances of deep convolutional neural networks. However, the direct migration of existing methods to video is still difficult to achieve good performance due to its lack of alignment and consistency modelling in temporal domain. Taking advantage of high inter-frame dependency in videos, we propose a self-enhanced convolutional network for facial video hallucination. It is implemented by making full usage of preceding super-resolved frames and a temporal window of adjacent low-resolution frames. Specifically, the algorithm first obtains the initial high-resolution inference of each frame by taking into consideration a sequence of consecutive low-resolution inputs through temporal consistency modelling. It further recurrently exploits the reconstructed results and intermediate features of a sequence of preceding frames to improve the initial super-resolution of the current frame by modelling the coherence of structural facial features across frames. Quantitative and qualitative evaluations demonstrate the superiority of the proposed algorithm against state-of-the-art methods. Moreover, our algorithm also achieves excellent performance in the task of general video super-resolution in a single-shot setting.
This paper studies face recognition (FR) and normalization in surveillance imagery. Surveillance FR is a challenging problem that has great values in law enforcement. Despite recent progress in conventional FR, less effort has been devoted to surveillance FR. To bridge this gap, we propose a Feature Adaptation Network (FAN) to jointly perform surveillance FR and normalization. Our face normalization mainly acts on the aspect of image resolution, closely related to face super-resolution. However, previous face super-resolution methods require paired training data with pixel-to-pixel correspondence, which is typically unavailable between real-world low-resolution and high-resolution faces. FAN can leverage both paired and unpaired data as we disentangle the features into identity and non-identity components and adapt the distribution of the identity features, which breaks the limit of current face super-resolution methods. We further propose a random scale augmentation scheme to learn resolution robust identity features, with advantages over previous fixed scale augmentation. Extensive experiments on LFW, WIDER FACE, QUML-SurvFace and SCface datasets have shown the effectiveness of our method on surveillance FR and normalization.