Real time object tracking using fused color and edge cues

Abstract

In this paper, we present an intelligent fusion framework to combine two complementary features for visual tracking. Color histogram has attracted much attention in recent years due to its simplicity and robustness. The capability of color histogram to deal with partial occlusion and non-rigidity has been demonstrated. In practice, however, relying on color information only is not sufficient in cases like scene illumination variations and background clutter. In order to perform tracking more reliably, edge orientation histogram (EOH), a feature that is invariant to illumination changes, discriminative against a background with confusing colors, and simple and fast to compute, is also employed. The proposed fusion framework is intelligent in that it adaptively combines the results from the two representation modules, where the more reliable cue always provides more weight on the final decision. The algorithm is tested on extensive real-world sequences and shown to achieve robust and reliable real-time tracking.