Reproduce “Progressive Color Transfer with Dense Semantic Correspondences” 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
Mingming He, Jing Liao, Dongdong Chen, Lu Yuan, Pedro V. Sander
Researchers from Hong Kong University of Science and Technology, City University of Hong Kong, and Microsoft Research introduced a progressive, coarse-to-fine framework that iteratively refines dense semantic correspondences and local color transfer. This approach enables accurate object-to-object color transformations, reduces artifacts, and supports multi-reference style guidance for images with diverse content.
Key insights
- Jointly optimizing semantic correspondences and color transfer is essential for preserving fine details and achieving semantically accurate transfers, avoiding artifacts seen in decoupled methods.
- Multi-level deep features are robust for establishing meaningful correspondences even when source and reference images have large appearance variations.
- A local color transfer model must enforce both local smoothness and non-local consistency to yield spatially adaptive yet globally coherent results.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.