17 Jun
2010
17 Jun
'10
7:28 p.m.
On Thu, 17 Jun 2010, ydegoyon@gmail.com wrote:
face recognition is not based on colour in any sense, it is based on number and orientation of segments of the contours, at least in haar's algorithm ...
Those lines are expressed using pixels and each of those pixels have a colour.
A segment detector could detect lines all over a picture, but it needs to be thresholded to lines significant enough. Lines that are too subtle have to be left out. In dark faces, or à contre-jour, how does Haar's algorithm figure out that lines inside of the face are more significant than those outside ?
_ _ __ ___ _____ ________ _____________ _____________________ ... | Mathieu Bouchard, Montréal, Québec. téléphone: +1.514.383.3801