Abstract—In this paper, we propose a new method of using local color information for text localization in natural images. The FAST salient point detector was applied in the first step to detect salient points from natural images. Then, the local color information of each salient point was extracted. The salient points was transformed into a bi-modal space to discover the text property of these points. After that, a clustering algorithm was used to find clusters of text area on images and got the possible positions of the text area. The distribution density was estimated and the average distance of salient points was calculated on the possible text area. Finally, minimum bounding box of the text area was defined to represent the text region. Experimental results have shown the advantages and effectiveness of the proposed method in the text detection in natural images.
 

Index Terms—feature point detection, text detection, natural images

Cite: Chien-Cheng Lee and Shang-Fei Shen, "Text Localization in Natural Images Using Discriminative Local Color Information," International Journal of Signal Processing Systems, Vol. 5, No. 3, pp. 89-93, September 2017. doi: 10.18178/ijsps.5.3.89-93