Text detection in natural scene images with feature combination

In this paper, we proposed a method for text detection in natural scene images by feature combination under a coarse-to-fine framework. Firstly, color feature is used to segment images into color-uniform regions by a clustering algorithm. Then edge features are extracted to construct a weak classifier to classify the regions into candidates or background. After a layout analysis procedure, candidate regions are connected into text lines. Finally texture features, color features, and statistic OCR (Optical Character Reader) features are extracted to discriminate text/non-text with a support vector machine (SVM). Experimental results on a large dataset show that the combination of features in different detection stages is competent for text detection task.