TextSR: Diffusion Super-Resolution with Multilingual OCR Guidance

TL;DR

TextSR is a multimodal diffusion model that improves multilingual scene text image super-resolution by integrating OCR-guided text priors, leading to more accurate and legible text reconstruction in challenging images.

Contribution

The paper introduces TextSR, a novel diffusion-based super-resolution model that incorporates OCR and text priors to enhance multilingual scene text image quality.

Findings

Outperforms existing methods on TextZoom and TextVQA datasets

Effectively localizes text regions and models multilingual character shapes

Enhances text legibility and reduces hallucinated textures

Abstract

While recent advancements in Image Super-Resolution (SR) using diffusion models have shown promise in improving overall image quality, their application to scene text images has revealed limitations. These models often struggle with accurate text region localization and fail to effectively model image and multilingual character-to-shape priors. This leads to inconsistencies, the generation of hallucinated textures, and a decrease in the perceived quality of the super-resolved text. To address these issues, we introduce TextSR, a multimodal diffusion model specifically designed for Multilingual Scene Text Image Super-Resolution. TextSR leverages a text detector to pinpoint text regions within an image and then employs Optical Character Recognition (OCR) to extract multilingual text from these areas. The extracted text characters are then transformed into visual shapes using a UTF-8 based text encoder and cross-attention. Recognizing that OCR may sometimes produce inaccurate results in real-world scenarios, we have developed two innovative methods to enhance the robustness of our model. By integrating text character priors with the low-resolution text images, our model effectively guides the super-resolution process, enhancing fine details within the text and improving overall legibility. The superior performance of our model on both the TextZoom and TextVQA datasets sets a new benchmark for STISR, underscoring the efficacy of our approach.