Reproduce “Fast Context-Based Low-Light Image Enhancement via Neural Implicit Representations” 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
Tomáš Chobola, Yu Liu, Hanyi Zhang, Julia A. Schnabel, Tingying Peng
CoLIE leverages neural implicit representations to estimate illumination in the HSV color space, enabling zero-shot low-light image enhancement that improves detail recovery and color fidelity, and enhances performance for downstream tasks such as object detection on high-resolution images.
Key insights
- Neural implicit representations can effectively learn detailed illumination maps for low-light image enhancement, moving beyond their typical use in pre-processing.
- Integrating local context information (pixel patches) as input to the implicit function significantly enriches the learned representation, leading to more robust illumination estimation.
- Operating in the HSV color space and focusing enhancement on the Value component successfully disentangles brightness manipulation from color, preserving original hues and saturations.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.