Skip to Main Content
Text localization is an important task in text understanding systems. In this paper, an efficient connected component based text localization method is proposed. Two image cues, edge and color, are combined to extract candidate text regions. The non-text regions in these candidate text regions are then removed by the text line information generated from 2D tensor voting. The images in ICDAR 2003 competition dataset are used to evaluate the proposed method. The experimental results show that the proposed method extracts more text regions with a low false positive rate.