Real-time finger tracking with improved performance in video sequences with motion blur

In this paper we propose a new algorithm for finger detection and tracking in video sequences captured with a common webcam. The proposed method uses a multi cue approach to achieve finger detection. Background subtraction and skin color segmentation are used prior to generating a sparse image representation based on line strip features. A multilayer processing is applied to the line strips in order to obtain quasi continuous clusters that can be identified as finger candidates. The tracking algorithm was tested in various challenging situations showing outstanding performance with motion blurred images and good performances with scale changes and occlusion handling. The proposed algorithm has a computationally effective architecture and is able to run in realtime.