ITR/SY: Software Systems for Vision-Based Spatial Interaction

Johns Hopkins University, Baltimore MD

Investigators

Abstract

The proposed work focuses on development of modular adaptable software for vision-based human-computer interaction. The key idea is to use local visual interaction cues on a video stream - mono or stereo, 2D or 3D - shared between the user and computer. The use of such cues has a number of advantages, such as limiting the processing to only those parts of the image where interaction takes place, allowing image processing to be dynamically driven by the interaction, and an easily parameterized modular basis for software development. Questions of how visual interaction cues can be flexibly defined and robustly detected based on local informaiton will be explored.