Tutorial on minimum output sum of squared error filter
Date
2016
Authors
Sidhu, Rumpal Kaur, author
Draper, Bruce A., advisor
Beveridge, Ross J., committee member
Troup, Lucy, committee member
Journal Title
Journal ISSN
Volume Title
Abstract
One of the major challenges in computer vision is visual tracking. Many different techniques have been developed for visual tracking over time. Visual tracking using correlation filters is faster than other techniques. A visual tracking technique using Optimized Correlation Output Filters (OCOF), Average of Synthetic Exact Filter (ASEF) and Minimum Output Sum of Squared Error (MOSSE) filters is presented in this tutorial. The MOSSE filter is a stable correlation filter which can be initialized on a single frame of a video. The MOSSE filter adapts with the changes in the appearance of the object while tracking. Tracking using the MOSSE filter is not dependent on changes in lighting, non-rigid transformations, pose, and scale. This work presents systematic steps towards building the filter and its application as a tracker in visual tracking. A number of experiments are performed to demonstrate the working of the tracker.