Video Super-Resolution With Convolutional Neural Networks

By Armin Kappeler, Seunghwan Yoo, Qiqin Dai, Aggelos K. Katsaggelos. IEEE Transactions on Computational Imaging, 2016

626 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/795bc037f00e532c0fe6a404d08c5576c7aa9554

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

Citations in the corpus

TDAN: Temporally-Deformable Alignment Network for Video Super-Resolution (2018, methodology): “Degradation Methods: We compare our TDAN-based network with current state-of-the-art VSR and SISR networks: VSRnet [10], ESPCN [23], VESCPN [3], Liu et al .”
Detail-Revealing Deep Video Super-Resolution (2017, methodology): “[14] estimated optical flow and selected corresponding patches across frames to train a CNN.”
Real-Time Video Super-Resolution with Spatio-Temporal Networks and Motion Compensation (2016, methodology): “This design has been studied for video classiﬁcation and action recognition [23, 37], and was also one of the architectures proposed in VSRnet [22].”
Deep Learning for Image Super-Resolution: A Survey (2019, methodology): “…based on optical ﬂow algorithms, Liao et al. [182] employ various optical ﬂow methods to generate HR candidates and ensemble them by CNNs. VSRnet [183] and CVSRnet [184] implement motion compensation by Druleas algorithm [185], and uses CNNs to take successive frames as input and predict HR…”
Artificial intelligence in the creative industries: a review (2020, methodology): “ross frames as well as local spatial correlations within them. Early contributions applying deep learning to achieve video SR gathered multiple frames into a 3D volume which formed the input to a CNN [92]. Later work exploited temporal correlation via a motion compensation process before concatenati”
Frame-Recurrent Video Super-Resolution (2018, methodology): “[18] warp video frames I LR t − 1 and I LR t +1 onto the frame I LR t using the optical ﬂow method of Drulea and Nedevschi [7], concatenate the three frames and pass them through a convolutional neural network that produces the output frame I est t .”
NTIRE 2019 Challenge on Video Deblurring and Super-Resolution: Dataset and Study (2019, background): “[11] presents CNN architectures with early fusion with concatenated input frames or extracted features from the frames.”
Deep Video Deblurring for Hand-Held Cameras (2017, methodology): “Super resolution networks have been applied to video sequences before [12, 16, 40], but these approaches”
Convolutional Neural Networks for Inverse Problems in Imaging: A Review (2017, methodology): “For video super-resolution, [18] improves on non-CNN-based methods by about 0.5 dB PSNR and [20] improves upon that result by another 0.5 dB.”
VRT: A Video Restoration Transformer (2022, methodology): “Early methods [5, 35, 51, 54, 88] use traditional flow estimation methods to estimate optical flow and warp the supporting frames towards the reference frame.”
EDVR: Video Restoration With Enhanced Deformable Convolutional Networks (2019, methodology): “Earlier studies [36, 4, 33, 19, 11] treat video restoration as a simple extenBicubic RCAN”
Progressive Fusion Video Super-Resolution Network via Exploiting Non-Local Spatio-Temporal Correlations (2019, methodology): “With recent development of deep learning, a lot of video SR methods [3, 14, 23, 26, 31] based on convolutional neural networks (CNNs) have emerged.”
Using Deep Neural Networks for Inverse Problems in Imaging: Beyond Analytical Methods (2018, methodology): “[17] (2016) CS Linear mapping followed by six-layer CNN; takes CS measurements as input and outputs an intermediate reconstruction, which is then fed to a denoiser”
Video Enhancement with Task-Oriented Flow (2017, background): “ng, researchers have directly modeled video enhancement tasks as regression problems without representing motions, and have designed deep networks for frame interpolation [21,24] and super-resolution [14,10,12,29]. Recently, with differentiable image sampling layers in deep learning [11], motion inf”
Recurrent Back-Projection Network for Video Super-Resolution (2019, background): “The most popular approach to retain temporal information in VSR is by concatenating the frames as in [18, 2, 16, 22].”
Deep Back-Projection Networks for Super-Resolution (2018, background): “Significant progress in deep learning for vision [15, 13, 5, 39, 26, 33, 17] has recently been propagating to the field of super-resolution (SR) [19, 29, 6, 12, 20, 21, 24, 42].”
Deep Video Super-Resolution Network Using Dynamic Upsampling Filters Without Explicit Motion Compensation (2018, methodology): “Upscale Metric Bicubic VSRnet[16] VESPCN[1] Tao et al.”
Multi-frame Quality Enhancement for Compressed Video (2018, methodology): “proposed a Video Super-Resolution network (VSRnet) [20], in which the neighboring frames are warped according to the estimated motion, and both the current and warped neighboring frames are fed into a super-resolution CNN to enlarge the resolution of the current frame.”