Contextual Point Matching for Video Stabilization: Sea Shell Sketch
Authors:Joseph Geumlek, Ling Han Meng
Mentor:Ernie Esser, Postdoctoral Scholar, University of California Irvine
We explore the potential of applying a particular shape matching algorithm to the domain of video stabilization. We wish to discover whether this algorithm can compete with existing video stabilization methods, and whether it excels for specific video data where other methods fail. This method is a natural fit for finding the point correspondences between subsequent frames in a video. By using global contextual information, this method out performs methods which only consider local features in cases where the shapes involved have high degrees of self-similarity. Though computationally-intensive, we found it to provide better results than existing methods without significantly increasing computational costs. Furthermore, this method performs well on sketching videos where popular methods fail.