*A curated list of resources for Image and Video Deblurring*
https://github.com/subeeshvasu/Awesome-Deblurring
GitHub - subeeshvasu/Awesome-Deblurring: A curated list of resources for Image and Video Deblurring
A curated list of resources for Image and Video Deblurring - GitHub - subeeshvasu/Awesome-Deblurring: A curated list of resources for Image and Video Deblurring
github.com
사용해본 Deblurring 기법
1. MTRNN(Multi-Temporal Recurrent Neural Networks), 2020
https://github.com/Dong1P/MTRNN
GitHub - Dong1P/MTRNN
Contribute to Dong1P/MTRNN development by creating an account on GitHub.
github.com
2. DeMFI(Deblurring and Multi Frame Interpolation), 2021
https://github.com/JihyongOh/DeMFI
GitHub - JihyongOh/DeMFI: Official repository of DeMFI (ECCV 2022).
Official repository of DeMFI (ECCV 2022). Contribute to JihyongOh/DeMFI development by creating an account on GitHub.
github.com
3. DeepDeblur(Deep Multi-scale Convolutional Neural Network for Dynamic Scene Deblurring), 2017
https://github.com/SeungjunNah/DeepDeblur-PyTorch
GitHub - SeungjunNah/DeepDeblur-PyTorch: Deep Multi-scale CNN for Dynamic Scene Deblurring
Deep Multi-scale CNN for Dynamic Scene Deblurring. Contribute to SeungjunNah/DeepDeblur-PyTorch development by creating an account on GitHub.
github.com
기회가 되면 각각의 기법 분석도 해보는걸로...
다음에 사용해보고픈 기법
RealTime_VDBLR(Real-Time Video Deblurring via Lightweight Motion Compensation), 2022 –
https://github.com/codeslake/RealTime_VDBLR
GitHub - codeslake/RealTime_VDBLR: [PG 2022] Official PyTorch Implementation for "Real-Time Video Deblurring via Lightweight Mot
[PG 2022] Official PyTorch Implementation for "Real-Time Video Deblurring via Lightweight Motion Compensation" - GitHub - codeslake/RealTime_VDBLR: [PG 2022] Official PyTorch Implementati...
github.com
BiT(Blur Interpolation Transformer), 2023 – https://github.com/zzh-tech/BiT
GitHub - zzh-tech/BiT: [CVPR2023] Blur Interpolation Transformer for Real-World Motion from Blur
[CVPR2023] Blur Interpolation Transformer for Real-World Motion from Blur - GitHub - zzh-tech/BiT: [CVPR2023] Blur Interpolation Transformer for Real-World Motion from Blur
github.com
실험 1. GoPro Dataset
정량적인 평가에서는 DeepDeblur가 속도 측면에서는 압도적이었고, PSNR, SSIM의 경우에도 약간 우수한 결과를 보였다.
하지만, 정성적인 평가(육안으로 비교)에서는 "택배" 글씨가 적힌 배너를 각각 비교하면서 보았을 때 deblurring 결과가 엄청 우수하지는 않아보였다.
실험 2. Custom Dataset
약 1~2초 영상을 blur가 심하도록 과격히 움직이고, 이를 fps=30으로 프레임을 쪼개 이미지로 구성하였다.
정량적인 평가에서는 MTRNN이 우수했으나, 정성적인 평가를 했을 때, DeepDeblur-PyTorch가 속도도 빠를 뿐더러 deblurring이 가장 준수하게 되었다.