A novel similar background components connection algorithm for colorful text detection in natural images

In this paper, a novel similar background components connection algorithm is presented to detect colorful texts. Before, the components were extracted accurately according to a color-based stroke width extraction rule proposed in the paper. In the stage of filtering non-text components, a specially designed metric is presented to adjust the parameter values of the filter, so that the filter is more robust under various conditions. Experiments on the ICDAR 2003 test set demonstrate that the proposed algorithm outperforms the state-of-the-art methods when detecting texts in same color. Furthermore, it achieves significantly performance on detecting colorful texts in natural images.