Purchase this article with an account.
W. I. A. Al-Atabany, T. S. Tong, P. Degenaar; A Novel Content Aware Scene Retargeting for Retinitis Pigmentosa Patients. Invest. Ophthalmol. Vis. Sci. 2010;51(13):1353.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
In this study we present a novel scene retargeting technique to shrink the visual scene into the remaining field of view for Retinitis Pigmentosa (RP) patients. Traditional simple compression makes objects appear further away. Our method compresses the visual scene into the remaining visual field of the patients while maintaining the size of the important features. The algorithm has been developed and tested experimentally and will be tested with patient trials to examine its functionality and efficiency.
Our method is a combination of the best aspects of two previous techniques to retarget images non-linearly. In summary, to achieve compression, our technique creates an importance map of each frame. The algorithm then determines a dynamic shrinking of the scene, shrinking important features lightly and less important features strongly.Currently, we have implemented our algorithm on a standard desktop computer for the initial patient trials. Because our algorithm is scalable, our next phase is to implement the algorithm on a portable electronic device. This will be connected to a virtual reality headset display. Our first patient trial phase includes testing the performance of the proposed algorithm by projecting recorded dynamic scene movies before and after compression onto a screen. Patient questions include the counting of the number of occurrences of certain events in each case. Efficacy is determined both quantitatively and qualitatively.
Our algorithm has been evaluated to other scene retargeting approaches from a content preservation and a compression quality perspective. Results, on recorded video files, show the robustness of our approach compared to the other techniques in preserving the scene contents intact and in the compressibility performance
Our developed algorithm can dynamically compress video sequences without knowledge of future frames, keeping the important objects relatively intact. The subsequent video stream has sufficiently low jitter to be acceptable to our application in patients with tunnel vision.
This PDF is available to Subscribers Only