Character Proposal Network for Robust Text Extraction

TL;DR

This paper introduces a character proposal network (CPN) that leverages fully convolutional networks to improve text detection by predicting character scores and refining locations, outperforming traditional MSER methods especially in challenging cases.

Contribution

The novel CPN combines high-capacity FCNs with a multi-template strategy to enhance character proposal accuracy and robustness in scene text detection.

Findings

Achieves over 93% recall on multiple datasets

Outperforms MSER in challenging scenarios

Uses fewer than 1000 proposals for high accuracy

Abstract

Maximally stable extremal regions (MSER), which is a popular method to generate character proposals/candidates, has shown superior performance in scene text detection. However, the pixel-level operation limits its capability for handling some challenging cases (e.g., multiple connected characters, separated parts of one character and non-uniform illumination). To better tackle these cases, we design a character proposal network (CPN) by taking advantage of the high capacity and fast computing of fully convolutional network (FCN). Specifically, the network simultaneously predicts characterness scores and refines the corresponding locations. The characterness scores can be used for proposal ranking to reject non-character proposals and the refining process aims to obtain the more accurate locations. Furthermore, considering the situation that different characters have different aspect ratios, we propose a multi-template strategy, designing a refiner for each aspect ratio. The extensive experiments indicate our method achieves recall rates of 93.88%, 93.60% and 96.46% on ICDAR 2013, SVT and Chinese2k datasets respectively using less than 1000 proposals, demonstrating promising performance of our character proposal network.