Automatic Image Processing and Conversion to Tactile Graphics

Graphical information has become a critical method for portraying information for education, work and personal tasks and decisions. Unfortunately there are currently limited means of providing this information to individuals who are blind or visually impaired: alternate text is frequently missing, and accessible tactile diagrams tend to be time consuming to make and require expertise in order for them to be interpretable (which may be costly to the user and/or impossible to get). The aim of this project is to provide an accessible system to automatically generate tactile graphics for those who need to interpret information contained in visual images. Previous automatic conversion methods have not been especially successful and are not used in normal practice, possibly because they have not taken advantage of current advances in the field of image processing. In the preliminary work, we systematically look at the myriad of image segmentation methods that exist as part of the conversion process. For those techniques, previous researchers have often compared the results to the “gold standard” of human segmentation to evaluate their success. However, there are important difference between this “gold standard” and what is needed for tactile graphics. Key steps by professionals who create tactile diagrams are simplification so that the information is manageable to extract through the tactile sense, elimination of perspective as it is difficult to interpret tactually, and possible spreading of information across multiple diagrams. Planned work is to examine more closely the underlying themes to the myriad of algorithm are relevant for tactile diagrams. Future work, will also involve taking the initially segmented image, simplifying it further by removing “unimportant” detail so that it is manageable by the tactile system and removing perspective based on geometric information found in the image