Tracking of objects in spatiotemporal volume by graph-cuts

Tracking of objects in the spatiotemporal volume is a recent approach to solving the tracking problem. Graph cut based tools exist for tracking objects in the spatiotemporal volume, but these algorithms are still in the rudimentary stage and they require user interaction for successful tracking. In this paper an attempt to automate the tracking process is done by choosing appropriate energy functions and using local optical flow data with background subtraction techniques. The energy functions used and their respective weight assignments are presented. The algorithm developed is tested with a synthetic sequence and a real sequence. Experimental results show excellent tracking for the synthetic sequence and reasonable tracking for the real sequence.