Template matching based video tracking system using a novel N-step search algorithm and HOG features

Abstract

A novel video object tracking technique is proposed in this article. We consider a robust template-matching based video tracking technique that works satisfactory for both static-camera and moving-camera video sequences, being not influenced by the camera motions. In our approach, the first instance of the video object is selected interactively. Then, its successive instances in the video frames are detected using a novel and improved N-step search algorithm for motion estimation taking into account both the scaling and translation of the target. A HOG-based feature extraction approach is used by our algorithm. © 2012 Springer-Verlag.