The Self Affine Feature Transform is capable to extract various geometric information
from geometric
image details, and also to organize these kind of details into di#?#erent classes.
This article shows how
di#?#erent details can be normalized through the normalization of the quadratic function
describing the
accumulated
ow velocity. The presented algorithm is independent of image contents (image class),
but
the uncertainty of the resulting normalizing transformation depends on which class
the image belongs.
A sample code is given to ease reproduction of results. The algorithm does not take
into account how
affine transformation a#?#ects low pass #?#ltering. E#?#ects on windowing is partly
incorporated.
