Text Localization, Extraction and Inpainting in Color Images using Combined Structural and Textural Features

In this article, a new approach is proposed for text detection, extraction and inpainting in color images. The proposed algorithm includes three stages. In the first stage, several gradient based operators and image corners are utilized to localize text blocks. We use a new block split and merging algorithm to enhance the accuracy of text localization algorithm. An SVM based text verification algorithm is then employed with a new set of features to reject non-text blocks. To inpaint text areas, we cluster different colors in the text blocks using k-means clustering algorithm and estimate background and text colors. Then a color segmentation algorithm is employed to detect characters’ pixels accurately. In the third stage, the proposed inpainting algorithm is applied to restore initial image contents. The inpainting algorithm is based on a matching algorithm that considers the priority for inpainting the pixels. Experimental results and a comparison of the results with those of other methods show the efficiency of the proposed algorithm.

Keywords:

text detection , text localization , image inpainting , structural and textural features , color segmentation