Improving Diffusion Models for Scene Text Editing with Dual Encoders

TL;DR

This paper introduces DIFFSTE, a dual encoder diffusion model that significantly improves scene text editing by enhancing text accuracy and style control, with strong zero-shot generalization capabilities demonstrated across multiple datasets.

Contribution

The paper proposes a novel dual encoder diffusion framework with instruction tuning, enabling better text rendering, style control, and zero-shot generalization in scene text editing.

Findings

Outperforms previous methods in text correctness and naturalness

Achieves effective style control and zero-shot font variation generation

Demonstrates superior results on five benchmark datasets

Abstract

Scene text editing is a challenging task that involves modifying or inserting specified texts in an image while maintaining its natural and realistic appearance. Most previous approaches to this task rely on style-transfer models that crop out text regions and feed them into image transfer models, such as GANs. However, these methods are limited in their ability to change text style and are unable to insert texts into images. Recent advances in diffusion models have shown promise in overcoming these limitations with text-conditional image editing. However, our empirical analysis reveals that state-of-the-art diffusion models struggle with rendering correct text and controlling text style. To address these problems, we propose DIFFSTE to improve pre-trained diffusion models with a dual encoder design, which includes a character encoder for better text legibility and an instruction encoder for better style control. An instruction tuning framework is introduced to train our model to learn the mapping from the text instruction to the corresponding image with either the specified style or the style of the surrounding texts in the background. Such a training method further brings our method the zero-shot generalization ability to the following three scenarios: generating text with unseen font variation, e.g., italic and bold, mixing different fonts to construct a new font, and using more relaxed forms of natural language as the instructions to guide the generation task. We evaluate our approach on five datasets and demonstrate its superior performance in terms of text correctness, image naturalness, and style controllability. Our code is publicly available. https://github.com/UCSB-NLP-Chang/DiffSTE