Real-time content-aware video retargeting on the Android platform for tunnel vision assistance

Show full item record

Redirect: RIT Scholars content from RIT Digital Media Library has moved from to RIT Scholar Works, please update your feeds & links!
Title: Real-time content-aware video retargeting on the Android platform for tunnel vision assistance
Author: Knack, Thomas
Abstract: As mobile devices continue to rise in popularity, advances in overall mobile device processing power lead to further expansion of their capabilities. This, coupled with the fact that many people suffer from low vision, leaves substantial room for advancing mobile development for low vision assistance. Computer vision is capable of assisting and accommodating individuals with blind spots or tunnel vision by extracting the necessary information and presenting it to the user in a manner they are able to visualize. Such a system would enable individuals with low vision to function with greater ease. Additionally, offering assistance on a mobile platform allows greater access. The objective of this thesis is to develop a computer vision application for low vision assistance on the Android mobile device platform. Specifically, the goal of the application is to reduce the effects tunnel vision inflicts on individuals. This is accomplished by providing an in-depth real-time video retargeting model that builds upon previous works and applications. Seam carving is a content-aware retargeting operator which defines 8-connected paths, or seams, of pixels. The optimality of these seams is based on a specific energy function. Discrete removal of these seams permits changes in the aspect ratio while simultaneously preserving important regions. The video retargeting model incorporates spatial and temporal considerations to provide effective image and video retargeting. Data reduction techniques are utilized in order to generate an efficient model. Additionally, a minimalistic multi-operator approach is constructed to diminish the disadvantages experienced by individual operators. In the event automated techniques fail, interactive options are provided that allow for user intervention. Evaluation of the application and its video retargeting model is based on its comparison to existing standard algorithms and its ability to extend itself to real-time. Performance metrics are obtained for both PC environments and mobile device platforms for comparison.
Record URI:
Date: 2012-05

Files in this item

Files Size Format View
TKnackThesis5-2012.pdf 18.29Mb PDF View/Open

The following license files are associated with this item:

This item appears in the following Collection(s)

Show full item record

Search RIT DML

Advanced Search