A pattern recognition approach to make accessible the geographic images for blind and visually impaired

In this paper, we propose an intelligent detection and recognition of the different geographic forms and the text presented in geographic images. The aim is to automate the process of converting these images into information that could be easily understandable by the visually impaired. The presented work includes the proposition of various appropriate pattern recognition techniques. A new method of detecting text in images is proposed. It is based on Curvelet and Stroke Width transforms. Moreover, a fuzzy segmentation step is performed in order to identify and extract the different geographic areas in the image; an automatic recognition algorithm of geometric symbols is also presented. The obtained results are very encouraging and promising.