Reproduce “Burstormer: Burst Image Restoration and Enhancement Transformer” 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

Akshay Dudhane, Syed Waqas Zamir, Salman Khan, Fahad Shahbaz Khan, Ming-Hsuan Yang

Burstormer introduces a transformer-based architecture for burst image restoration and enhancement, processing multiple degraded RAW frames into a single high-quality sRGB image. The model achieves state-of-the-art performance across burst super-resolution, low-light enhancement, and denoising tasks, often with reduced parameters and faster execution.

Key insights

  • Transformers, with their self-attention mechanisms, are effective for capturing long-range dependencies and global context crucial for alignment and robust feature fusion in burst images.
  • Progressive multi-scale alignment from coarse to fine resolutions, combined with extensive reference-based feature enrichment, enhances robustness to complex and residual misalignments.
  • Adaptive pooling for variable burst sizes and cyclic burst sampling for progressive, long-range inter-frame communication enable flexible and efficient information aggregation.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.