TransientAreasSegmentationModule |
class which provides a transient/moving areas segmentation module
perform a locally adapted segmentation by using the retina magno input data Based on Alexandre
BENOIT thesis: "Le système visuel humain au secours de la vision par ordinateur"
3 spatio temporal filters are used:
a first one which filters the noise and local variations of the input motion energy
a second (more powerfull low pass spatial filter) which gives the neighborhood motion energy the
segmentation consists in the comparison of these both outputs, if the local motion energy is higher
to the neighborhood otion energy, then the area is considered as moving and is segmented
a stronger third low pass filter helps decision by providing a smooth information about the
"motion context" in a wider area
|