Background
Prime video silhouette tracking is based on the ability to track objects under unlimited conditions. Since vast majority of trackers rely on limited models, tracking objects under general conditions such as non-rigid targets whose appearances may change over time, general camera motion or a 3D scene may fail.
Moreover, even the latest visual trackers rely on the assumption of which the targeted object constantly fits thse model in terms of shape and appearances, but when it cease to obey the model - the tracker is likely to fail. 

Method
Our innovative Bittracker track objects by applying two stages:
• Approximating, in each frame, a probability distribution function (PDF) of the targeted object bitmap.
• Estimating the maximum a posteriori bitmap.

Since the PDF function is calculated over all possible pixels per motion it is possible to avoid the optical flow stagnation stage.
Moreover, the Bittracker model is built only on basic assumptions whereas specific object attributes such as motion or scene or even a priori information do not affect the tracker.
As a result, it is possible to track silhouettes in video streams under general conditions.

Advantages
• Suitable for tracking under very general conditions
• Operating without a priori object related information

Applications
• Alpha- matting process in video editing
• Security cameras