Section: New Results

High tech vision aid-systems for low-vision patients

Improving social interaction through augmented reality

Participants : Josselin Gautier, Pierre Kornprobst, Nicolas Chleq, Frédéric Dosière [Bosch Visiontec (Sophia Antipolis, France)] , David Coupé [Bosch Visiontec (Sophia Antipolis, France)] .

Today’s visual enhancement systems for low-vision people consist of dedicated augmented reality hardware allowing to magnify or enhance the overall scene, independently of the image content or patient needs. For example, for patients with central vision loss, interacting with others may become a painful activity when faces and expressions can hardly be recognized. In [17], we introduce a new augmented reality system allowing to selectively enhance faces, using two image processing techniques  [44], [50]. Our system is based on a Fove 0 head-mounted display (FOVE Inc, San Mateo, CA, USA). It has the capacity to adjust the enhancement to the detected faces’ size and distance, hence maintaining a constant boost in the critical range of spatial frequency. It offers a binocular and large Field-of-View and performs at near real-time with a modest laptop computer using multithreading. Preliminary experiments with three patients with central vision loss suggest that the enhancements chosen strongly depends on each patient’s condition and lead to improved recognition abilities when patients find their optimal settings.

This work is presented in [17].

Text auto-illustration for improving reading accessibility to low-vision people

Participants : Paula Pawlowski, Pierre Kornprobst, Elena Cabrio [Inria, EPI WIMMICS] , Marco Benzi [Université Côte d'Azur (France)] .

We have started to explore how to make reading more efficient and more enjoyable for low-vision patients through text auto-illustration. Text auto-illustration consists in automatically extracting images from the web which are related to a given text, using natural language processing methods.