TextCohesion: Detecting Text for Arbitrary Shapes

TL;DR

TextCohesion is a novel pixel-wise scene text detection method that decomposes text into key components, effectively handling complex backgrounds and curved texts, outperforming existing methods on challenging benchmarks.

Contribution

The paper introduces a new pixel-wise approach with a confidence scoring mechanism for improved scene text detection, especially for arbitrary shapes.

Findings

Achieves 84.6% F-measure on Total-Text

Achieves 86.3% F-measure on SCUT-CTW1500

Outperforms state-of-the-art methods on curved text benchmarks

Abstract

In this paper, we propose a pixel-wise method named TextCohesion for scene text detection, which splits a text instance into five key components: a Text Skeleton and four Directional Pixel Regions. These components are easier to handle than the entire text instance. A confidence scoring mechanism is designed to filter characters that are similar to text. Our method can integrate text contexts intensively when backgrounds are complex. Experiments on two curved challenging benchmarks demonstrate that TextCohesion outperforms state-of-the-art methods, achieving the F-measure of 84.6% on Total-Text and bfseries86.3% on SCUT-CTW1500.