Reproduce “Collaborative Feedback Discriminative Propagation for Video Super-Resolution” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Hao Li, Xiang Chen, Jiangxin Dong, Jinhui Tang, Jinshan Pan
This research introduces Collaborative Feedback Discriminative (CFD) propagation for video super-resolution, which enhances reconstruction by adaptively correcting inaccurately aligned features and enabling richer spatio-temporal information exploration. Integrated into existing VSR backbones, the method achieves up to 0.42 dB PSNR improvement on Vimeo-T and competitive state-of-the-art results on REDS4, demonstrating clearer image details and structural integrity.
Key insights
- Leveraging shallow features extracted directly from the current low-resolution frame can effectively guide the correction of high-frequency and structural information degraded by inaccurate motion alignment.
- A bidirectional feedback mechanism, incorporating hidden states from both forward and backward propagation branches along with previous feedback features, significantly enriches temporal information interaction and refinement.
- Gated collaborative feed-forward blocks enhance spatio-temporal modeling by allowing features to learn from different timesteps through self-gating and interaction with future propagation features.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.