A New Approach for Text Location Based on SUSAN and SVM

Abstract

In order to extract text parts from images, this paper proposes a new approach that combines edge detection, heuristic knowledge and SVM. Firstly, SUSAN algorithm with an adaptive threshold is used to extract the edge information from gray images, and the binary edge images are obtained; Secondly, the heuristic knowledge is used to filter the text candidate areas obtained by edge dilation, and the positions of the text candidate areas in the binary image apply to the gray image; Finally we use statistics of gray level co-occurrence matrix as SVM's inputs to construct SVM which is used to verify the text areas further. The experimental results show that the proposed approach can effectively locate the text areas with high accurate and extraction ratios.