Deep Video Deblurring for Hand-Held Cameras

By Shuochen Su, Mauricio Delbracio, Jue Wang, Guillermo Sapiro, Wolfgang Heidrich, Oliver Wang. 2017

610 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Media Technology

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

In Media Technology

Citations in the corpus

The Unreasonable Effectiveness of Deep Features as a Perceptual Metric (2018, methodology): “Video deblurring We sample from the video deblurring dataset [52], along with deblurring outputs from Photoshop Shake Reduction, Weighted Fourier Aggregation [11], and three variants of a deep video deblurring method [52].”
Artificial intelligence in the creative industries: a review (2020, methodology): “with fewer ghosting artefacts compared to previous methods. In the case of video sequences, deblurring can benet from the abundant information present across neighbouring frames. The DeBlurNet model [109] takes a stack of nearby frames as input and uses synthetic motion blur to generate a training d”
Real-World Blur Dataset for Learning and Benchmarking Deblurring Algorithms (2020, background): “Recently, several synthetic datasets for learning image deblurring have been proposed [31, 30, 32, 39, 51].”
Adversarial Spatio-Temporal Learning for Video Deblurring (2018, methodology): “Different from FCNN and DBN [48], spatial”
Deep Image Deblurring: A Survey (2022, methodology): “In addition, 2D convolutions are also adopted by all video deblurring methods (Aittala and Durand 2018; HyunKim et al. 2017;Kim et al. 2018;Nah et al. 2019, ?; Sim and Kim 2019; Su et al. 2017;Wang et al. 2019;Wieschollek et al. 2017).”
Human-Aware Motion Deblurring (2019, methodology): “Data Collection: Following the non-uniform blur generation methodology proposed in [37, 46], we capture videos at 240fps with a GoPro Hero camera.”
DeblurGAN-v2: Deblurring (Orders-of-Magnitude) Faster and Better (2019, methodology): “We take 600 images from GoPRO, and 600 images from DVD, both with motion blurs already (same as above).”
Spatio-Temporal Filter Adaptive Network for Video Deblurring (2019, methodology): “For a fair comparison with the most related deep learning-based algorithms [23, 17, 44, 38], we finetune all these methods by the corresponding publicly released implementations on the video deblurring dataset [36].”
v2e: From Video Frames to Realistic DVS Events (2021, methodology): “To better estimate ﬂow for luma frames, we retrained Super-SloMo on the Adobe240FPS [32] dataset after converting its RGB frames to luma frames.”
Deep Stacked Hierarchical Multi-Patch Network for Image Deblurring (2019, methodology): “[21] presented a deep learning framework to process blurry video sequences and accumulate information across frames.”
NTIRE 2019 Challenge on Video Deblurring and Super-Resolution: Dataset and Study (2019, methodology): “4, compared with linear [28] and gamma function [23] that are assumed in blur synthesis.”
Patch-Based Image Inpainting with Generative Adversarial Networks (2018, background): “Recent advances in generative models show that deep neural networks can synthesize realistic looking images remarkably, in applications such as super-resolution [15, 18, 6], deblurring [28], denoising [39] and inpainting [25, 34, 11, 21].”
Deblurring by Realistic Blurring (2020, neutral): “, [21, 33, 36, 42].”
VRT: A Video Restoration Transformer (2022, methodology): “C. Video Deblurring 1) Quantitative Results: We conducts experiments on three different datasets for fair comparison with existing meth-ods [2], [4], [5], [6], [16], [22], [24], [36], [62], [64].”
Deblurring via Stochastic Refinement (2021, methodology): “Alternatively, many recent methods adopt an end-to-end approach where a deep neural network is trained to directly produce a point estimate [8, 11, 14, 21, 35, 36, 50, 54, 55, 64, 66, 67, 73].”
EDVR: Video Restoration With Enhanced Deformable Convolutional Networks (2019, methodology): “Several approaches [34, 50] directly fuse multiple frames without explicit temporal alignment, because the existence of blur increases the difficulty of motion estimation.”