Efficient Scene-Text Refinement via Pixel-Space Grouped Image Generation

Hugging Face Daily Papers Papers
scene-text-refinement pixel-space-generation image-editing mean-flow ocr text-generation

Summary

PixText 将局部文本修正建模为像素空间的分组图像生成，仅生成目标区域并保留未编辑内容，借助 MeanFlow 实现四步快速生成，在新提出的 TextRefinementBench 上以 83.06% 的字形通过率大幅超越最强基线。

Generated images often have coherent layouts and visual content, yet local errors in small text and complex characters still limit their readability and practical use. Scene-text refinement requires accuracy, preservation, and efficiency. Existing methods based on variational autoencoders (VAEs) can lose fine strokes during reconstruction, while full-image editing can shift background colors. In this work, we introduce PixText, a model that formulates local text refinement as grouped image generation in pixel space. Conditioned on the original image, PixText generates only the target regions, preserves unedited content, and uses MeanFlow for four-step generation. We further introduce Text Refinement Bench to evaluate text correction and image preservation in generated images. On this benchmark, PixText achieves 83.06% glyph pass and 94.66% exact target-character OCR accuracy, versus 36.77% and 52.31% for the strongest baseline. Its four-step MeanFlow variant retains an 81.84% pass rate with less than one GPU-second of amortized generation cost per image.
Original Article
View Cached Full Text

Cached at: 10/15/26, 07:04 AM

Paper page - Efficient Scene-Text Refinement via Pixel-Space Grouped Image Generation

Source: https://huggingface.co/papers/2610.18284

Abstract

Generatedimagesoftenhavecoherentlayoutsandvisualcontent,yetlocalerrorsinsmalltextandcomplexcharactersstilllimittheirreadabilityandpracticaluse.Scene-textrefinementrequiresaccuracy,preservation,andefficiency.Existingmethodsbasedonvariationalautoencoders(VAEs)canlosefinestrokesduringreconstruction,whilefull-imageeditingcanshiftbackgroundcolors.Inthiswork,weintroducePixText,amodelthatformulateslocaltextrefinementasgroupedimagegenerationinpixelspace.Conditionedontheoriginalimage,PixTextgeneratesonlythetargetregions,preservesuneditedcontent,andusesMeanFlowforfour-stepgeneration.WefurtherintroduceTextRefinementBenchtoevaluatetextcorrectionandimagepreservationingeneratedimages.Onthisbenchmark,PixTextachieves83.06%glyphpassand94.66%exacttarget-characterOCRaccuracy,versus36.77%and52.31%forthestrongestbaseline.Itsfour-stepMeanFlowvariantretainsan81.84%passratewithlessthanoneGPU-secondofamortizedgenerationcostperimage.

View arXiv pageView PDFAdd to collection

Get this paper in your agent:

hf papers read 2610\.18284

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2610.18284 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2610.18284 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2610.18284 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

On the Design Fundamentals of Pixel Text Representation Learning

Hugging Face Daily Papers

Pixel Linguist II improves visual text representation learning via variable resolution training, natural image-text grounding, layout-aware rendering, and multilingual curricula, achieving state-of-the-art results and robust compression performance.

UltraText Bench: A Comprehensive Bilingual Benchmark for Evaluating Visual Text Rendering in Image Generation

Hugging Face Daily Papers

UltraText Bench is a bilingual (English/Chinese) benchmark with 432 prompts across 24 scene categories for evaluating dense visual text rendering in image generation models, scored by the Q-Judge VLM across text fidelity, clarity, spatial quality, and scene quality — revealing strong trade-offs between model variants such as Z-Image-Turbo versus Z-Image-Base.