Text Localization in Unconstrained Images

Text localization is the first step when automatically reading text in images. Since existing methods often fail when applied to unconstrained images, in this paper we propose a more robust approach exploiting different kind of information. In particular, we first extract textural features, a combination of a stroke filter with a super-pixel segmentation, and then search for connected components. To finally obtain a text localization, these are subsequently analyzed for unary character properties, binary character similarities, and text line properties. To demonstrate the benefits of the proposed method, we evaluate it on three different data sets, showing