On the other hand, let’s consider the case where some OHL pixels are occupied by a stationary obstructing foreground and hence cannot detect the moving object at any time. An everyday analogy would be a camera tracking a car that passes by and disappears behind another vehicle parked in the foreground. Those pixels seeing the foreground never had the opportunity to learn the car image and hence will not be able to recognize it. This situation is simulated by a 2 x 2 pixel white square travelling from left to right while occluded along the way by a stationary 2 x 2 pixel red square, as illustrated in movie 3, with Figures 5a-e mirroring snapshots of this event. For the image frames in which occlusion takes place (Figures 5b to d), those OHL pixels representing the masked portion of the object have never seen and hence were not able to learn these hidden parts and are hence not able to reconstruct them. To manage such a scenario, path prediction is a viable method which can be performed by utilizing the OHL pixels’ learning function. That is, tracking can be achieved by forecasting the route the obstructed object will most likely take based on its previous trajectory. Our human brain works in a similar fashion, anticipating an occluded object to reappear following its past motion, based on experience. It is a natural assumption we subconsciously make. To foresee the most likely trail, the OHL pixels should be able to detect and to learn the direction and speed of the path taken by the object.