Abstract:
This paper takes place in the field of handwriting analysis.
More precisely, our aim is to extract some information
from the writing image itself. An OCR could improve its
recognition rate from this information. We are concerned
with 300 dpi off line text, and are working on the binary
images of the text lines. We are establishing the definition
of some new parameters that rely on the evolution of line
profiles when changing observation scale. We show how
these parameters can quantify such properties of
handwritten texts as the degree of loops on stroke or the
average leaning, for example. Also, some experiments
have been pursued on printed texts in order to have a
reference.