1 | initial version |
If the head is the only moving object in the scene, then you will get good results with background-foreground segmentation followed by tracking of blobs that occur.
Another approach could be to create a multiview detector specified for a human head. An example of this can be found in this paper.