f***@beethoven.com
2005-05-23 22:07:25 UTC
Given a database of images of faces, an algorythm to identify
similarities is given by applying morphing transformations between two
pairs of images for a number of intermediate images. The differential
between the number of points changed would give a measure of the
diffrence between two faces, for a normalized grayscale image. Also a
quantization with octries can help approximate similarities for an
appropriate gradation of grays...
These algorythms work even for small regions of the image if they are
normalized by size too. Of course, a conventional framework for the
inpout of face images will help as a preprocessor for the processing of
the images. Through this method a distribution of facial features can
be found and a measure of distinctiveness between human beings, similar
to the differences in genotypes. It can be complemented with a
psychological research measuring the sensitive to differences between
individuals (twins, for instance) in person and/or pictures...
Using an automatic classification algorythm (kohonen), characterologies
and typologies can be generated, to be correlated to other
characteristics compiled for each individual. This approach can serve
for tasks such as law making, career selection, prediction, marketing,
etc...
Fabrizio J Bonsignore, now Danilo J Bonsignore
similarities is given by applying morphing transformations between two
pairs of images for a number of intermediate images. The differential
between the number of points changed would give a measure of the
diffrence between two faces, for a normalized grayscale image. Also a
quantization with octries can help approximate similarities for an
appropriate gradation of grays...
These algorythms work even for small regions of the image if they are
normalized by size too. Of course, a conventional framework for the
inpout of face images will help as a preprocessor for the processing of
the images. Through this method a distribution of facial features can
be found and a measure of distinctiveness between human beings, similar
to the differences in genotypes. It can be complemented with a
psychological research measuring the sensitive to differences between
individuals (twins, for instance) in person and/or pictures...
Using an automatic classification algorythm (kohonen), characterologies
and typologies can be generated, to be correlated to other
characteristics compiled for each individual. This approach can serve
for tasks such as law making, career selection, prediction, marketing,
etc...
Fabrizio J Bonsignore, now Danilo J Bonsignore