Reproduce “Screentone-Preserved Manga Retargeting” 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
Minshan Xie, Menghan Xia, Xueting Liu, Tien-Tsin Wong
A learning-based approach addresses the challenge of resizing manga images by preserving the integrity of screentones, utilizing a "copy-and-paste" strategy to select and arrange patterns from the original image. This method achieves superior visual coherence and quantitative results (e.g., higher PSNR, SSIM, lower LPIPS) compared to conventional retargeting and synthesis-based techniques.
Key insights
- Screentones are best preserved by selectively borrowing and arranging original patterns rather than through interpolation or synthesis.
- Manga screentones exhibit perceptual invariance to small translations, which can be captured by a specialized loss function ($L_{sis}$).
- An attention-driven, recurrent selection mechanism can effectively integrate hierarchical multi-scale screentone proposals guided by semantic information.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.