StatLUT framework enables text-driven photorealistic style transfer

Researchers have introduced StatLUT, a novel multimodal framework for generating 3D Look-Up Tables (LUTs) for photorealistic style transfer. This framework addresses limitations in existing deep learning methods by decoupling color style from structural semantics using a Lab-Extractor, thus preventing unnatural distortions. StatLUT formulates LUT generation as a Transformer-based Seq2Seq translation task with a Multi-dimensional Residual Mapper to ensure topologically smooth LUTs. Additionally, it incorporates H-Diffuser, a Diffusion Transformer that enables text-driven color grading by synthesizing features from natural language prompts. AI

IMPACT This research could lead to more intuitive and artifact-free image style transfer, potentially impacting creative tools and media production.

RANK_REASON The cluster contains a research paper detailing a new framework for image style transfer.

Read on arXiv cs.CV →

• alphaXiv
• arXiv
• CatalyzeX Code Finder for Papers
• Connected Papers
• CORE Recommender
• DagsHub
• Diffusion Transformer
• Gotit.pub
• H-Diffuser
• Hugging Face
• Influence Flower
• Lab-Extractor
• Litmaps
• Multi-dimensional Residual Mapper
• ScienceCast
• scite Smart Citations
• StatLUT
• Transformer

• paper
• product

AI-generated summary · Google Gemini · from 2 sources. How we write summaries →