Guiding visual attention in grayscale images for auditory substitution systems

Abstract

Image-to-sound substitution systems aim to provide mobility assistance for visually disabled persons by exploiting the auditory augmentation. In general the existing auditory substitution systems, due to the processing limitations, discard the color and employ mostly grayscale image versions. We introduce a novel decolorization startegy inspired from the ability of the human visual system to automatically attend to the salient regions of complex scenes. This characteristic enables our nervous system to allocate limited perceptual and cognitive resources on relevant visual input. Our method guided by the salient map performs fast being able to preserve the original appearance as well as in the decolorized image version. The results demonstrate the utility of our technique to improve the quality of rendition in comparison with related approaches.