Video super-resolution (VSR) methods have recently achieved a remarkable success due to the development of deep convolutional neural networks (CNN). Current state-of-the-art CNN methods usually treat the VSR problem as a large number of separate multi-frame super-resolution tasks, at which a batch of low resolution (LR) frames is utilized to generate a single high resolution (HR) frame, and running a slide window to select LR frames over the entire video would obtain a series of HR frames. However, duo to the complex temporal dependency between frames, with the number of LR input frames increase, the performance of the reconstructed HR frames become worse. The reason is in that these methods lack the ability to model complex temporal dependencies and hard to give an accurate motion estimation and compensation for VSR process. Which makes the performance degrade drastically when the motion in frames is complex. In this paper, we propose a Motion-Adaptive Feedback Cell (MAFC), a simple but effective block, which can efficiently capture the motion compensation and feed it back to the network in an adaptive way. Our approach efficiently utilizes the information of the inter-frame motion, the dependence of the network on motion estimation and compensation method can be avoid. In addition, benefiting from the excellent nature of MAFC, the network can achieve better performance in the case of extremely complex motion scenarios. Extensive evaluations and comparisons validate the strengths of our approach, and the experimental results demonstrated that the proposed framework is outperform the state-of-the-art methods.
Video super-resolution (SR) aims at generating a sequence of high-resolution (HR) frames with plausible and temporally consistent details from their low-resolution (LR) counterparts. The key challenge for video SR lies in the effective exploitation of temporal dependency between consecutive frames. Existing deep learning based methods commonly estimate optical flows between LR frames to provide temporal dependency. However, the resolution conflict between LR optical flows and HR outputs hinders the recovery of fine details. In this paper, we propose an end-to-end video SR network to super-resolve both optical flows and images. Optical flow SR from LR frames provides accurate temporal dependency and ultimately improves video SR performance. Specifically, we first propose an optical flow reconstruction network (OFRnet) to infer HR optical flows in a coarse-to-fine manner. Then, motion compensation is performed using HR optical flows to encode temporal dependency. Finally, compensated LR inputs are fed to a super-resolution network (SRnet) to generate SR results. Extensive experiments have been conducted to demonstrate the effectiveness of HR optical flows for SR performance improvement. Comparative results on the Vid4 and DAVIS-10 datasets show that our network achieves the state-of-the-art performance.
Video super-resolution aims at generating a high-resolution video from its low-resolution counterpart. With the rapid rise of deep learning, many recently proposed video super-resolution methods use convolutional neural networks in conjunction with explicit motion compensation to capitalize on statistical dependencies within and across low-resolution frames. Two common issues of such methods are noteworthy. Firstly, the quality of the final reconstructed HR video is often very sensitive to the accuracy of motion estimation. Secondly, the warp grid needed for motion compensation, which is specified by the two flow maps delineating pixel displacements in horizontal and vertical directions, tends to introduce additional errors and jeopardize the temporal consistency across video frames. To address these issues, we propose a novel dynamic local filter network to perform implicit motion estimation and compensation by employing, via locally connected layers, sample-specific and position-specific dynamic local filters that are tailored to the target pixels. We also propose a global refinement network based on ResBlock and autoencoder structures to exploit non-local correlations and enhance the spatial consistency of super-resolved frames. The experimental results demonstrate that the proposed method outperforms the state-of-the-art, and validate its strength in terms of local transformation handling, temporal consistency as well as edge sharpness.
While Convolutional Neural Networks (CNNs) trained for image and video super-resolution (SR) regularly achieve new state-of-the-art performance, they also suffer from significant drawbacks. One of their limitations is their lack of robustness to unseen image formation models during training. Other limitations include the generation of artifacts and hallucinated content when training Generative Adversarial Networks (GANs) for SR. While the Deep Learning literature focuses on presenting new training schemes and settings to resolve these various issues, we show that one can avoid training and correct for SR results with a fully self-supervised fine-tuning approach. More specifically, at test time, given an image and its known image formation model, we fine-tune the parameters of the trained network and iteratively update them using a data fidelity loss. We apply our fine-tuning algorithm on multiple image and video SR CNNs and show that it can successfully correct for a sub-optimal SR solution by entirely relying on internal learning at test time. We apply our method on the problem of fine-tuning for unseen image formation models and on removal of artifacts introduced by GANs.
Super-resolution (SR) has been widely used to convert low-resolution legacy videos to high-resolution (HR) ones, to suit the increasing resolution of displays (e.g. UHD TVs). However, it becomes easier for humans to notice motion artifacts (e.g. motion judder) in HR videos being rendered on larger-sized display devices. Thus, broadcasting standards support higher frame rates for UHD (Ultra High Definition) videos (4K@60 fps, 8K@120 fps), meaning that applying SR only is insufficient to produce genuine high quality videos. Hence, to up-convert legacy videos for realistic applications, not only SR but also video frame interpolation (VFI) is necessitated. In this paper, we first propose a joint VFI-SR framework for up-scaling the spatio-temporal resolution of videos from 2K 30 fps to 4K 60 fps. For this, we propose a novel training scheme with a multi-scale temporal loss that imposes temporal regularization on the input video sequence, which can be applied to any general video-related task. The proposed structure is analyzed in depth with extensive experiments.
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.
We propose a novel method that tracks fast moving objects, mainly non-uniform spherical, in full 6 degrees of freedom, estimating simultaneously their 3D motion trajectory, 3D pose and object appearance changes with a time step that is a fraction of the video frame exposure time. The sub-frame object localization and appearance estimation allows realistic temporal super-resolution and precise shape estimation. The method, called TbD-3D (Tracking by Deblatting in 3D) relies on a novel reconstruction algorithm which solves a piece-wise deblurring and matting problem. The 3D rotation is estimated by minimizing the reprojection error. As a second contribution, we present a new challenging dataset with fast moving objects that change their appearance and distance to the camera. High speed camera recordings with zero lag between frame exposures were used to generate videos with different frame rates annotated with ground-truth trajectory and pose.
For video super-resolution, current state-of-the-art approaches either process multiple low-resolution (LR) frames to produce each output high-resolution (HR) frame separately in a sliding window fashion or recurrently exploit the previously estimated HR frames to super-resolve the following frame. The main weaknesses of these approaches are: 1) separately generating each output frame may obtain high-quality HR estimates while resulting in unsatisfactory flickering artifacts, and 2) combining previously generated HR frames can produce temporally consistent results in the case of short information flow, but it will cause significant jitter and jagged artifacts because the previous super-resolving errors are constantly accumulated to the subsequent frames. In this paper, we propose a fully end-to-end trainable frame and feature-context video super-resolution (FFCVSR) network that consists of two key sub-networks: local network and context network, where the first one explicitly utilizes a sequence of consecutive LR frames to generate local feature and local SR frame, and the other combines the outputs of local network and the previously estimated HR frames and features to super-resolve the subsequent frame. Our approach takes full advantage of the inter-frame information from multiple LR frames and the context information from previously predicted HR frames, producing temporally consistent high-quality results while maintaining real-time speed by directly reusing previous features and frames. Extensive evaluations and comparisons demonstrate that our approach produces state-of-the-art results on a standard benchmark dataset, with advantages in terms of accuracy, efficiency, and visual quality over the existing approaches.
The video super-resolution (VSR) task aims to restore a high-resolution (HR) video frame by using its corresponding low-resolution (LR) frame and multiple neighboring frames. At present, many deep learning-based VSR methods rely on optical flow to perform frame alignment. The final recovery results will be greatly affected by the accuracy of optical flow. However, optical flow estimation cannot be completely accurate, and there are always some errors. In this paper, we propose a novel deformable non-local network (DNLN) which is a non-optical-flow-based method. Specifically, we apply the deformable convolution and improve its ability of adaptive alignment at the feature level. Furthermore, we utilize a non-local structure to capture the global correlation between the reference frame and the aligned neighboring frames, and simultaneously enhance desired fine details in the aligned frames. To reconstruct the final high-quality HR video frames, we use residual in residual dense blocks to take full advantage of the hierarchical features. Experimental results on benchmark datasets demonstrate that the proposed DNLN can achieve state-of-the-art performance on VSR task.
Video super-resolution reconstruction (SRR) algorithms attempt to reconstruct high-resolution (HR) video sequences from low-resolution observations. Although recent progress in video SRR has significantly improved the quality of the reconstructed HR sequences, it remains challenging to design SRR algorithms that achieve good quality and robustness at a small computational complexity, being thus suitable for online applications. In this paper, we propose a new adaptive video SRR algorithm that achieves state-of-the-art performance at a very small computational cost. Using a nonlinear cost function constructed considering characteristics of typical innovation outliers in natural image sequences and an edge-preserving regularization strategy, we achieve state-of-the-art reconstructed image quality and robustness. This cost function is optimized using a specific alternating projections strategy over non-convex sets that is able to converge in a very few iterations. An accurate and very efficient approximation for the projection operations is also obtained using tools from multidimensional multirate signal processing. This solves the slow convergence issue of stochastic gradient-based methods while keeping a small computational complexity. Simulation results with both synthetic and real image sequences show that the performance of the proposed algorithm is similar or better than state-of-the-art SRR algorithms, while requiring only a small fraction of their computational cost.
The popularity of high and ultra-high definition displays has led to the need for methods to improve the quality of videos already obtained at much lower resolutions. Current Video Super-Resolution methods are not robust to mismatch between training and testing degradation models since they are trained against a single degradation model (usually bicubic downsampling). This causes their performance to deteriorate in real-life applications. At the same time, the use of only the Mean Squared Error during learning causes the resulting images to be too smooth. In this work we propose a new Convolutional Neural Network for video super resolution which is robust to multiple degradation models. During training, which is performed on a large dataset of scenes with slow and fast motions, it uses the pseudo-inverse image formation model as part of the network architecture in conjunction with perceptual losses, in addition to a smoothness constraint that eliminates the artifacts originating from these perceptual losses. The experimental validation shows that our approach outperforms current state-of-the-art methods and is robust to multiple degradations.
Objects moving at high speed along complex trajectories often appear in videos, especially videos of sports. Such objects elapse non-negligible distance during exposure time of a single frame and therefore their position in the frame is not well defined. They appear as semi-transparent streaks due to the motion blur and cannot be reliably tracked by standard trackers. We propose a novel approach called Tracking by Deblatting based on the observation that motion blur is directly related to the intra-frame trajectory of an object. Blur is estimated by solving two intertwined inverse problems, blind deblurring and image matting, which we call deblatting. The trajectory is then estimated by fitting a piecewise quadratic curve, which models physically justifiable trajectories. As a result, tracked objects are precisely localized with higher temporal resolution than by conventional trackers. The proposed TbD tracker was evaluated on a newly created dataset of videos with ground truth obtained by a high-speed camera using a novel Trajectory-IoU metric that generalizes the traditional Intersection over Union and measures the accuracy of the intra-frame trajectory. The proposed method outperforms baseline both in recall and trajectory accuracy.
Video restoration tasks, including super-resolution, deblurring, etc, are drawing increasing attention in the computer vision community. A challenging benchmark named REDS is released in the NTIRE19 Challenge. This new benchmark challenges existing methods from two aspects: (1) how to align multiple frames given large motions, and (2) how to effectively fuse different frames with diverse motion and blur. In this work, we propose a novel Video Restoration framework with Enhanced Deformable networks, termed EDVR, to address these challenges. First, to handle large motions, we devise a Pyramid, Cascading and Deformable (PCD) alignment module, in which frame alignment is done at the feature level using deformable convolutions in a coarse-to-fine manner. Second, we propose a Temporal and Spatial Attention (TSA) fusion module, in which attention is applied both temporally and spatially, so as to emphasize important features for subsequent restoration. Thanks to these modules, our EDVR wins the champions and outperforms the second place by a large margin in all four tracks in the NTIRE19 video restoration and enhancement challenges. EDVR also demonstrates superior performance to state-of-the-art published methods on video super-resolution and deblurring. The code is available at https://github.com/xinntao/EDVR.
Recently, image super-resolution has been widely studied and achieved significant progress by leveraging the power of deep convolutional neural networks. However, there has been limited advancement in video super-resolution (VSR) due to the complex temporal patterns in videos. In this paper, we investigate how to adapt state-of-the-art methods of image super-resolution for video super-resolution. The proposed adapting method is straightforward. The information among successive frames is well exploited, while the overhead on the original image super-resolution method is negligible. Furthermore, we propose a learning-based method to ensemble the outputs from multiple super-resolution models. Our methods show superior performance and rank second in the NTIRE2019 Video Super-Resolution Challenge Track 1.
Recently, deep learning based video super-resolution (SR) methods have achieved promising performance. To simultaneously exploit the spatial and temporal information of videos, employing 3-dimensional (3D) convolutions is a natural approach. However, straight utilizing 3D convolutions may lead to an excessively high computational complexity which restricts the depth of video SR models and thus undermine the performance. In this paper, we present a novel fast spatio-temporal residual network (FSTRN) to adopt 3D convolutions for the video SR task in order to enhance the performance while maintaining a low computational load. Specifically, we propose a fast spatio-temporal residual block (FRB) that divide each 3D filter to the product of two 3D filters, which have considerably lower dimensions. Furthermore, we design a cross-space residual learning that directly links the low-resolution space and the high-resolution space, which can greatly relieve the computational burden on the feature fusion and up-scaling parts. Extensive evaluations and comparisons on benchmark datasets validate the strengths of the proposed approach and demonstrate that the proposed network significantly outperforms the current state-of-the-art methods.
We proposed a novel architecture for the problem of video super-resolution. We integrate spatial and temporal contexts from continuous video frames using a recurrent encoder-decoder module, that fuses multi-frame information with the more traditional, single frame super-resolution path for the target frame. In contrast to most prior work where frames are pooled together by stacking or warping, our model, the Recurrent Back-Projection Network (RBPN) treats each context frame as a separate source of information. These sources are combined in an iterative refinement framework inspired by the idea of back-projection in multiple-image super-resolution. This is aided by explicitly representing estimated inter-frame motion with respect to the target, rather than explicitly aligning frames. We propose a new video super-resolution benchmark, allowing evaluation at a larger scale and considering videos in different motion regimes. Experimental results demonstrate that our RBPN is superior to existing methods on several datasets.
We study the video super-resolution (SR) problem for facilitating video analytics tasks, e.g. action recognition, instead of for visual quality. The popular action recognition methods based on convolutional networks, exemplified by two-stream networks, are not directly applicable on video of low spatial resolution. This can be remedied by performing video SR prior to recognition, which motivates us to improve the SR procedure for recognition accuracy. Tailored for two-stream action recognition networks, we propose two video SR methods for the spatial and temporal streams respectively. On the one hand, we observe that regions with action are more important to recognition, and we propose an optical-flow guided weighted mean-squared-error loss for our spatial-oriented SR (SoSR) network to emphasize the reconstruction of moving objects. On the other hand, we observe that existing video SR methods incur temporal discontinuity between frames, which also worsens the recognition accuracy, and we propose a siamese network for our temporal-oriented SR (ToSR) training that emphasizes the temporal continuity between consecutive frames. We perform experiments using two state-of-the-art action recognition networks and two well-known datasets--UCF101 and HMDB51. Results demonstrate the effectiveness of our proposed SoSR and ToSR in improving recognition accuracy.
In video super-resolution, the spatio-temporal coherence between, and among the frames must be exploited appropriately for accurate prediction of the high resolution frames. Although 2D convolutional neural networks (CNNs) are powerful in modelling images, 3D-CNNs are more suitable for spatio-temporal feature extraction as they can preserve temporal information. To this end, we propose an effective 3D-CNN for video super-resolution, called the 3DSRnet that does not require motion alignment as preprocessing. Our 3DSRnet maintains the temporal depth of spatio-temporal feature maps to maximally capture the temporally nonlinear characteristics between low and high resolution frames, and adopts residual learning in conjunction with the sub-pixel outputs. It outperforms the most state-of-the-art method with average 0.45 and 0.36 dB higher in PSNR for scales 3 and 4, respectively, in the Vidset4 benchmark. Our 3DSRnet first deals with the performance drop due to scene change, which is important in practice but has not been previously considered.
Video super-resolution (VSR) aims to restore a photo-realistic high-resolution (HR) video frame from both its corresponding low-resolution (LR) frame (reference frame) and multiple neighboring frames (supporting frames). Due to varying motion of cameras or objects, the reference frame and each support frame are not aligned. Therefore, temporal alignment is a challenging yet important problem for VSR. Previous VSR methods usually utilize optical flow between the reference frame and each supporting frame to wrap the supporting frame for temporal alignment. Therefore, the performance of these image-level wrapping-based models will highly depend on the prediction accuracy of optical flow, and inaccurate optical flow will lead to artifacts in the wrapped supporting frames, which also will be propagated into the reconstructed HR video frame. To overcome the limitation, in this paper, we propose a temporal deformable alignment network (TDAN) to adaptively align the reference frame and each supporting frame at the feature level without computing optical flow. The TDAN uses features from both the reference frame and each supporting frame to dynamically predict offsets of sampling convolution kernels. By using the corresponding kernels, TDAN transforms supporting frames to align with the reference frame. To predict the HR video frame, a reconstruction network taking aligned frames and the reference frame is utilized. Experimental results demonstrate the effectiveness of the proposed TDAN-based VSR model.
Our work explores temporal self-supervision for GAN-based video generation tasks. While adversarial training successfully yields generative models for a variety of areas, temporal relationships in the generated data are much less explored. Natural temporal changes are crucial for sequential generation tasks, e.g. video super-resolution and unpaired video translation. For the former, state-of-the-art methods often favor simpler norm losses such as $L^2$ over adversarial training. However, their averaging nature easily leads to temporally smooth results with an undesirable lack of spatial detail. For unpaired video translation, existing approaches modify the generator networks to form spatio-temporal cycle consistencies. In contrast, we focus on improving learning objectives and propose a temporally self-supervised algorithm. For both tasks, we show that temporal adversarial learning is key to achieving temporally coherent solutions without sacrificing spatial detail. We also propose a novel Ping-Pong loss to improve the long-term temporal consistency. It effectively prevents recurrent networks from accumulating artifacts temporally without depressing detailed features. Additionally, we propose a first set of metrics to quantitatively evaluate the accuracy as well as the perceptual quality of the temporal evolution. A series of user studies confirm the rankings computed with these metrics. Code, data, models, and results are provided at https://github.com/thunil/TecoGAN. The project page https://ge.in.tum.de/publications/2019-tecogan-chu/ contains supplemental materials.