Combining color and ViBe foreground mask for better tracking in video Sequences

Color histogram is an important visual cue for tracking targets of interest in video sequences. However, tracking using color is prone to errors due to: (i) background clutter (neighborhood regions having similar appearance with the target) and/or (ii) the target enters a shadowy areas. To overcome these two limitations; a tracker combining color and foreground mask is presented in this paper. The tracker is based on particle filter, where a joint 16×16×16-bins histogram built in the RGB color space is used as the color cue, while the foreground mask is generated by a foreground/background (FG/BG) subtraction algorithm chosen after an experimental evaluation of three recently developed algorithms. First, experiments have been done using the color cue alone, where the tracker drifts away and loses the target in this case. After that, the color is combined with the foreground mask by adopting a likelihood based on a product strategy. In this case, better and improved tracking results are reported.