Although fast and reliable, real-time template tracking using linear predictors requires a long training time. The lack of the ability to learn new templates online prevents their use in applications that require fast learning. This especially holds for applications running on low-end hardware such as mobile phones, and also for SLAM and SfM applications where the scene is not known a priori and multiple templates have to be added online. So far, linear predictors had to be either learned offine [1] or in an iterative manner by starting with a small sized template and growing it over time [2]. In this paper, we propose a fast and simple reformulation of the learning procedure that allows learning new linear predictors online. We performed an exhaustive evaluation of our approach and compared it to standard linear predictors [1] for template tracking as well as to an iterative learning approach [2] and demonstrated that our approach does not lose tracking performance and is more robust to noise, while allowing a much faster learning. In addition, we demonstrate the usefulness of the proposed approach in mobile phone applications and similar applications to SLAM, where multiple templates are being tracked simultaneously.     «

Although fast and reliable, real-time template tracking using linear predictors requires a long training time. The lack of the ability to learn new templates online prevents their use in applications that require fast learning. This especially holds for applications running on low-end hardware such as mobile phones, and also for SLAM and SfM applications where the scene is not known a priori and multiple templates have to be added online. So far, linear predictors had to be either learned offine...     »