Image Processing and Pattern Recognition Tools for the Automatic Image Transcription
The main objective of this work is to automate the conversion process of pedagogical images into information easily understandable by blind people and visually impaired people. This is performed by determining automatically the different areas of interest
- PDF / 686,620 Bytes
- 7 Pages / 439.37 x 666.142 pts Page_size
- 54 Downloads / 197 Views
Abstract. The main objective of this work is to automate the conversion process of pedagogical images into information easily understandable by blind people and visually impaired people. This is performed by determining automatically the different areas of interest in the image, then identify each region by assigning it a texture which will be for example transformed in relief. The text present in the image is also detected, recognized and transformed in accessible text (in Braille text or vocal message). The solution that we offer by this work is to provide a tool that tries to find automatically the principal information conveyed by the image and then transmit it to blind people. Keywords: Image accessibility Image transcription Text detection Image resolution enhancement Curvelet transform Stroke Width Transform (SWT) OCR Image segmentation
1 Introduction Image accessibility is an important area of research. Indeed, tactile images are widely used as an efficient accessibility tool to visually impaired people in many domains especially in schools and educational environments. The principal drawback of relief images is that blind people do not know always Braille. Moreover, the interactive map is a fairly simple and convenient way to access information for these people. A type of interactive map combines relief map and voice synthesis [1]. Users explore the relief image with their finger. At the same time a tablet below the image gives vocal explanation which has the same functionality as a legend. Another technique provides a novel opportunity to present an image to blind people; it is the touch screen with electro-vibration feedback [2]. With this electro-vibration feedback technique, we can simulate touch with different textures on the screen. Unfortunately, until now, the tactile image creation is done manually. With a computer, the designers use generally an image editing software to redraw an image from its original version. Given the current scientific research development in pattern recognition and image processing in general, we propose solutions to this problem. To transcribe an image, two important stages must be realized: (i) the text detection, recognition and transcription in Braille, (ii) the segmentation of the different image areas and the affiliation of textures. Generally, images like pedagogical images often need to be enlarged and enhanced for better OCR detection. Indeed, OCR does not work with small text and could give © Springer International Publishing Switzerland 2016 K. Miesenberger et al. (Eds.): ICCHP 2016, Part I, LNCS 9758, pp. 197–203, 2016. DOI: 10.1007/978-3-319-41264-1_26
198
Z. Haddad et al.
very poor results with extended texts of poor quality. For these reason, we propose a new image resolution enhancement method which improves the quality of super resolution image. The classical technique to enlarge an image is interpolation. The classical image interpolation techniques are nearest neighbor interpolation, bilinear interpolation, and bi-cubic interpolation. Further
Data Loading...
