Increasing the speed of Model-based detection and tracking through Wavelet transform

Methods for tracking an object are being categorized to four groups: region-based, model-based, feature-based and active contour. Each of these methods has its own advantage and disadvantages. Model-based is very accurate but its computational time is very high and this causes problems in tracking especially in real-time applications. In model-based method, firstly we must adapt an object to the different frames of the video and then try to track it. Here we have proposed a new method that reduces the time of tracking very tangibly and are tested in two human body databases (CASIA and Mocap) with remarkable improvement in results.