Improving Feature Tracking by Robust Points of interest Selection

This paper deals with robust point features selection for tracking. The aim is to identify unreliable features since the first frame so to track them in all the sequence. We extend a recent version of the well-known Kanade-Lucas-Tomasi tracker [18] by introducing an automatic scheme for rejecting spurious features. We employ a simple and efficient rejection rule based on grey levels co-occurrence entropy and show that its empirically assumptions are satisfied in the scenario of feature tracking. Experiments with real and synthetic images confirm that this approach makes better features tracking. We illustrate quantitatively the benefits introduced by the proposed algorithm.

• Type: Conference
• Year: 2001
• Where: VMV
• Authors: Chafik Kermad, Christophe Collewet
• Keywords: Levels Co-occurrence Entropy, Robust Point Features, VMV 2001, VMV 2008, Well-known Kanade-lucas-tomasi Tracker