ABSTRACT:

The augmentation of real video feeds with virtual elements is an important aspect of applications in areas such as real-time broadcasting and augmented reality. This project addresses the insertion and automated placement of graphical annotations (text/images) into video streams. Specifically, it focuses on the placement of annotations in "unimportant" or less interesting regions of the video imagery so that essential scene elements stay unoccluded. Vineet Thanedar is working towards building a framework and system prototype that enables the automated placement of arbitrary-sized annotations in either pre-processed or live videos in real-time. He approaches the problem from a human perceptual perspective by employing computer vision and image analysis techniques to search for suitable placement locations in the video over space and time. He presently identifies uniformity, motion, and clutter as Elementary Perceptual Descriptors and analyzes the video for these measures. He employs quantitative formulations for each of the perceptual properties and combines them to rate video regions for their placement suitability. Vineet intends to develop a generic system that can be easily applied in diverse application domains, such as augmented reality, Interactive TV, commercial advertising in telecasts, annotation of geographic map data, and interactive instructional/training videos.