Tech Tips: Seeing Pictures

Dan Comden, DO-IT staff

How do blind people "see" pictures? The same way many of them read text-through their fingertips. However, traditional creation of tactile graphics is a time-consuming and laborious process. Delivering images from textbooks to blind students has meant hours of editing and creating Braille labels to replace print text. Simple images may be created in ten or fifteen minutes; complex images may take many hours. The Tactile Graphics Project at the University of Washington, which is funded by the National Science Foundation (grant #IIS0415273) has created software called the Tactile Graphics Assistant (TGA) to greatly speed this process, achieving processing rates under five minutes per image.

The software is taught to recognize text within sample images. Rather than use Optical Character Recognition (OCR) technology, the program looks at font size, color, and shape to determine what is text and what is part of the drawing. OCR programs do a poor job at recognizing text that is part of or near an image. Machine learning is at the core of the TGA program and allows for batch processing, executing a series of non-interactive jobs all at one time, once the basics of recognition are learned.

The TGA software then removes the picture of the text, retaining the location information from where it was grabbed. It places the imaged text in a separate graphics file. This file can then be processed with standard OCR software with much greater accuracy. The user can then convert this recognized text to Braille, preparing it for reinsertion. Custom scripts, small software programs that function on the Internet, used in imaging software such as Adobe Illustratorâ„¢ can then use the location information (retained when the print text was removed) to reinsert the Braille in the appropriate position. Finally, the tactile specialist can then do any minor editing necessary before producing the graphic on an output device, like an embosser (a hardware device for "printing" a hard copy of a text document in Braille).

This exciting project is the result of a combined effort between a number of UW departments, including the Access Technology Lab, DO-IT, the Information School, and the Department Computer Science and Engineering. Dr. Richard Ladner, UW computer science professor, is the Principal Investigator. We are hoping that this process will be used in future production in tactile graphics shops around the country, resulting in greater access to graphic scientific and math information for students who are blind.

More information about the Tactile Graphics Project can be found at tactilegraphics.cs.washington.edu.