Shadow Removal
-
Where L ? corresponded objects in frame t-1, i ? un-corresponded objects in frame t and ? is the weight parameter
- Linear models are used to updated velocity and change in size, once correspondence is established.
- In case of occlusion
- Velocity parameter not updated
- Velocity used to predict centroid position.
- Predicted Centroid and Bounding Box positions constrained within the foreground regions.