This thesis introduces a finger tracker, called RapidVision tracker, for multitouch camera-based systems. Different techniques in image processing are examined and evaluated in order to find an optimal processing chain for the tracker. Out of the many different methods, we choose those which provide results as accurate as possible and are at the same time as simple as possible. This assures better computation time than complex techniques which is most significant especially for real-time applications. After reading this thesis the reader will have acquired knowledge about some basic steps in computer vision and object tracking, such as denoising techniques, image segmentation approaches like background subtraction, and edge detection. Since it is necessary to identify the object before tracking it, this thesis presents and discusses a few object detection techniques in detail. The final phase of the processing chain is the tracking of the object. Some approaches considered appropriate for our purpose and application environment are characterized. Since our multi-touch system applies more than one camera, we must consider if sequential processing is efficient enough for real-time application. To improve the tracker's efficiency parallel processing of the processing steps is applied whenever possible and reasonable. The final and optimal processing chain is then compared to two other state-of-theart trackers with regard to its suitability for real-time systems, especially with respect performance and accuracy. A graphical user interface (GUI) is also provided for setting the parameters of the different image processing algorithms.