Restore Text First, Enhance Image Later: Two-Stage Scene Text Image Super-Resolution with Glyph Structure Guidance
先恢复文本,后增强图像:基于字形结构引导的两阶段场景文本图像超分辨率
机构 * VCIP, CS, Nankai University(南开大学计算机科学与技术学院) ; vivo Mobile Communication Co. Ltd(vivo移动通信有限公司) ; SDS, The Chinese University of Hong Kong, Shenzhen(香港中文大学(深圳)信息科学与技术学院)
AI总结 TIGER通过两阶段框架先恢复文本再增强图像,利用字形结构引导实现场景文本图像超分辨率的高保真度与可读性提升。
Comments Project Page: https://tony-lowe.github.io/TIGER_project_page/ GitHub Repo: https://github.com/OpenVeraTeam/TiGeSR Huggingface Dataset: https://huggingface.co/datasets/mxluocv/UZ-ST Huggingface Weights: https://huggingface.co/mxluocv/TiGeSR