Natural scene text localization using edge color signature

Abstract

Localizing text regions in images taken from natural scenes is one of the challenging problems dueto variations in font, size, color and orientation of text. In this paper, we introduce a new concept socalled Edge Color Signature for localizing text regions in an image. This method is able to localizeboth Farsi and English texts. In the proposed method rst a pyramid using different scales of theinput image is created. Then for each level of the pyramid an edge map is extracted. Afterward,several geometric features are employed to lter out the non-text edges from the extracted edges.At this stage we describe an edge using colors of its neighboring pixels. We use the mean-Shiftalgorithm to obtain the color modes surrounding each edge pixel. Subsequently, the connectededge pixels with similar color signatures are clustered using Single-Linkage clustering algorithm toconstruct meaningful groups. Finally, each of the clusters is labeled as text or non-text using anMLP based cascade classi er. The proposed method has been evaluated on well-known ICDAR 2013and our Farsi dataset, the result is very promising.