Re: FAce comparison

[l3 <mail@mail.com>]

I was assuming that all faces are frontal facing.

If the orientation of tha faces are not similar, then a proper 
localization procedure is required. And that in itself is a pretty big 
research topic.



Roy Schestowitz wrote:
> l3 wrote:
>
>
> > By saying that "between two images", I'm assuming that the image
> > contains the face which you want to compare.
> >
> > If that's the case, the process is called normalisation, or face
> > localisation (although localisation usually refers to a larger class of
> > problems).
> >
> > If it's just normalisation, then use the distance between the eyes or
> > the triangle formed by the eyes, and nose. That said, an eye-detector,
> > nose-detector or both will have to be implemented.
> >
> >
> >
> > mixo wrote:
> >
> > > Hello,
> > >
> > > I would like to implement an algorithm on face comparison between two
> > > images. I Have some ideas how to do like geometrical comparison but
> > > some problem appears as the position of the face or other.
> > > somebody have an idea to resolve this problem ?
> > >
> > > Thanks all.
>
>
> It then leads to the question: how powerful an algorithm does he need?
> Solely measuring distances in this way is very simplistic. It doesn't
> account for image intensities and it neglects the mouth, ears, etc.
>
> Roy