Advances in Video Super-Resolution and Image Restoration

The field of video super-resolution and image restoration is rapidly advancing with the development of new methods and techniques. One of the key trends is the use of diffusion models, which have shown promising results in video super-resolution tasks. These models are able to effectively capture temporal dependencies and fine spatial details, leading to significant improvements in reconstruction quality. Another area of focus is the development of more efficient and scalable methods, which can handle long sequences and high-definition videos. Researchers are also exploring new evaluation metrics and methodologies to better assess the performance of these models. Notably, some papers have proposed innovative approaches to address the challenges of video super-resolution, such as patch-wise processing and autoregressive modeling. Overall, the field is moving towards more efficient, effective, and scalable methods for video super-resolution and image restoration. Noteworthy papers include: DeLiVR, which proposes a differential spatiotemporal Lie bias for efficient video deraining, and PatchVSR, which breaks video diffusion resolution limits with patch-wise video super-resolution. Additionally, InfVSR and LVTINO have also made significant contributions to the field, with InfVSR breaking length limits of generic video super-resolution and LVTINO proposing a latent video consistency inverse solver for high-definition video restoration.

Sources

DeLiVR: Differential Spatiotemporal Lie Bias for Efficient Video Deraining

Towards Redundancy Reduction in Diffusion Models for Efficient Video Super-Resolution

Asymmetric VAE for One-Step Video Super-Resolution Acceleration

A Data-Centric Perspective on the Influence of Image Data Quality in Machine Learning Models

PatchVSR: Breaking Video Diffusion Resolution Limits with Patch-wise Video Super-Resolution

Continuous Space-Time Video Super-Resolution with 3D Fourier Fields

Image-Difficulty-Aware Evaluation of Super-Resolution Models

NSARM: Next-Scale Autoregressive Modeling for Robust Real-World Image Super-Resolution

Gather-Scatter Mamba: Accelerating Propagation with Efficient State Space Model

InfVSR: Breaking Length Limits of Generic Video Super-Resolution

LVTINO: LAtent Video consisTency INverse sOlver for High Definition Video Restoration

Pure-Pass: Fine-Grained, Adaptive Masking for Dynamic Token-Mixing Routing in Lightweight Image Super-Resolution

Built with on top of