-
Notifications
You must be signed in to change notification settings - Fork 0
Chamfer
samirahmed edited this page Oct 9, 2012
·
2 revisions
The intuition behind the Chamfer is that it represents the distance to the closed feature Where a feature represents a notable change in the luminousity
Two vectors of chamfer distances are considered Similar, when determining the scalar dot product, is similar to the dot product with itself.
The Chamfer calculation, is defined as the
Chamfer(p) = min( abs(lastFeature.position-p.position) , abs(nextFeature.position-p.position) )
In implementation, the first and last pixel are always considered to be features
Chamfer Similarity can calculated in O(n) time with the fomula below,
Similarity = (c1 . c2)/ max( c1.c1,c2.c2)
Where the . represents the scalar dot product of the vectors being compares