Automatic Description of SVG Images for the Visually Impaired: A Gestaltic Approach

In this paper, a new approach to the automatic description of SVG images for the visually impaired based on Gestalt theory is presented, using levels with increased details. The description begins as a whole, considering relations of alignment, symmetry and group formation of the elements. Next, qualitative values for shape, locations, measures and colors are used to characterize each element. Finally, quantitative values are given for these properties. Tests carried out with users, compared automatic descriptions with human made. There were improvements of 9%, using automatic descriptions. Considering only the visually impaired, this figure rises to 18%.