Papers on line
All my papers, together with a list of the
most interesting ones, are detailed in my CV. Moreover, the heading
[http://cmm.ensmp.fr/bibliotheque.html]
of this website shows the class marks of these
documents at the School
of Mines library from
1992, and even before for some of them.
However, as these two lists do not give a
direct access to the texts themselves, I put on line most of my publications
from 1994, together with some technical reports, which can be downloaded from
this page. These articles, communications and technical reports deal with the
following themes :
Connectivity : 1995 IEEE paper [A65], written with Ph. Salembier,
describes the connected filters properties for numerical functions (pyramids,
commutativity, contours) for image compression. The
mathematical study [A74]enlarges the connection notion to complete
lattices, which had been introduced in 1989 in the set lattices
framework [06_2].
Two more simple presentations
[A75] and [C63] come back on the general case and
on the set one, and take up with the filtering again [A65]
Segmentation : In publication [A77], in collaboration with Ch. Ronse, we tried to extend to the numerical functions the
“geodesic” dilations, which had only been set wise oriented up to
that moment. It was a trial to link the connection and the segmentation of
numerical functions. Afterwards, I preferred the approach developed in the
note [NI212] and presented in both communications
[C71] (in French, published in « Traitement du Signal »
Vol. 20 n° 3) and [C75] (in English, cf. CIC 2002 congress proceedings). These beginnings
have led to the main text about this subject, i.e. the technical report [NI238]], which is accepted by JIMV for
publication in 2005. In
this document, segmentation is studied like the research of optimum
partitions according to connected criteria. This way of looking at things
turns out to be a solution to the paradox between segmentation, considered as
an optimum (therefore as a global result) and the fact that it often comes
among other operations. We will find other results about mixed partitions in [C81].
Interpolation : the following research line is
independent from the previous paragraphs, and
originates in the video images interpolations to be compressed.
Several quite close lines were also followed by S. Beucher,
J.R. Casas, Ph. Salembier
and F. Meyer. For my part, I proved that Hausdorff
metric accepts several geodesics). Some of them can be made symmetrical,
which leads to quite high quality interpolators, and which broaden to
numerical functions [C61].
With M. Iwanowski, we studied the extension to affine versions [C64] as well as to colour vector
functions [C62].
Some examples of
these interpolations can be seen [fly.avi] ; [multi.avi] ; [ren_morph_c.avi].
Colour : I had to face colour problems when I was
working on interpolation. At the same time, A. Hanbury
and I were wondering how we could construct morphological operations for data
defined on the unit circle, such as directions or colours (see [A79], published in IEEE in 2001). This drove us
to pay more attention the luminance/saturation/colour systems of polar
coordinates, such as HLS or HSV. And we realized they were contradictory with
Newton’s
disk famous experience, because of a wrong definition of saturation [NI230][C77][T38]. We proposed some alternatives (see [A81], published in IAS {Image Analysis and Stereology} in 2001, or the most detailed version [NI205], or [C68]). On that point, as some robotics staff had asked us to detect
reflections on colour pictures, J. Angulo and I
were brought to study again the various systems of polar coordinates for
colour, and to compare their results. The norm L_{1} based system proved to be very informative, as
the reflection regions are represented by straight lines in the Luminance /
Saturation histogram. This new step led me to propose an original model for light
space distribution [C82].
Finally, combining my
studies on maximum partition segmentation with the saturation function as
“colour measuring”, J. Angulo and I
have created a segmentation method for colour images, presented in English in
[C79], in French in [NI236] and in Spanish in [C76]Another possibility consists in
building a total order, named lexicographic, on colour lattices [C80]. An application of this method to
map segmentation is proposed in the paper [C81].
The reference [NI239] gives an
overview of the works on colour at CMM, from 1998 to 2003 (in French).
Finally, a series of examples is published in the book dedicated to G. Matheron (2005, in English) [CDL8].
3Dimensions morphology : In 1999, within a few months, two embryology
problems were submitted to me, from two totally independent directions, and
both used to involve 3dimensional structures. Dr Bertram (Melbourne University)
was aiming at giving a quantitative description of kidneys embryos
development which were growing invitro, and Dr. Staub
(CHU LariboisièreSaint Denis, Paris) was working
on the morphogenesis of long bones, such as the shinbone, for orthopedic
research. From a morphological viewpoint, both questions are based on
geodesic (i.e.
submitted to constraints) wave
front propagations within the threedimensional environment. Branchings,
bottle necks, extremities, etc can be detected and study this way …[A80] (French version [NI194]). Also, thanks to these studies,
I could extend a more formal research [C60] in which I had proposed the
cubeoctahedron as a better digital substitute to the sphere.
Viscous lattices and other theoretical studies : Propagations by Euclidean geodesic
wave fronts consider the singletons as the smallest moving particles. But
what if the smallest moving element is not the point anymore, but a small
ball which radius r>0, as if there were space quanta ?
This question is dealt with in the viscous lattices study [NI237] (accepted in
2005 for publication by JIMV). As we are mentioning theory, let us quote a
note that was written after J. Barrera’s work on the operators
decomposition. I added several constraints in order to channel this too large
problem [NI209]. On the other side, [NI216] is a research over Euclidean
functions classes, that are suitable to morphological operations (they must
make up complete lattices, closed under the multiplication by 1, and
continuous enough). As this study turns us towards equicontinuous
functions, I kept it up and created a probabilistic version of equicontinuous classes [C55].
Biological applications : There are many forays of mathematical
morphology into biology. Dozens could be mentioned :
for instance, I took part to a study over the early detection of transplanted
kidneys (Dr D. Seron’s team in Barcelona), which was
rewarded as the best paper for the year 1996 in “Kidney
International”. To mention more recent studies, publication [A82] in “Bioinformatics” in
2003 offers a methodology to read biochips, which was created by J. Angulo, and which is presently used as an international
reference (particularly in the USA) for genetic analysis. Another study, more
medicinerelated, was carried out with Prof. Z. Massy’s
team at Necker
Hospital, and deals
with the aorta calcification ([A83],
published in IAS in
2003). Finally, a larger study, carried out with Prof. Flandrin
(Necker Hospital, Paris),
led to an
automatic classification of lymphocytes. I select three communications,
written with J. Angulo, two in English,
dealing with the chromatin structure [C67] and cells recognition [C74], and another one in Spanish,
about telemedicine applications [C70].
Other Applications : Biomedical applications are not the only
ones the Center
of Mathematical Morphology
is involved in. At present, many other fields are opened, but more
particularly material sciences and multimedia. As an example, here are two
studies from these two themes. The first one [C65] was undertaken with M. Młynarczuk (Krakow),
and synthesizes petrography information obtained from the
polarizer. The second one, with E. Decencière and
M.E. Díaz (Valencia), allows us to cancel
vertical lines in old movies [NI187]]. Some of these applications tackle unusual physics fields and do not
always refer to models. The invited lecture [C66] to ICPR 2000 congress (Barcelona) wonders if
shape recognition may be considered as physics.
Courses : My courses are displayed as
transparencies or written texts. For the first ones, they are listed into the
“transparencies” heading. For the second ones, I have given access
to five of them from this page. The first one, [T34], is a general text over deterministic
morphological methods. It was written for the first francoscandinavian
mathematics school, that Prof. Ch. Kiselman (Uppsala University)
and I organised at Lake Erken (Sweden,
2002). This course includes a part of the material on 3D morphology from [T36] course, lectured at Bordeaux Stereology congress in 2002, but teh
Bordeaux
course is more specific to 3D. Both third course about random sets and
functions [T37], with its exercises [T37b], fourth one on colour [T39], lectured at Lille University
in 2005 are totally independent from the two first courses (i.e. Lake Erken
and Bordeaux).
Finally, reference [NI216]
that I mentioned
before, about models of numerical functions, is one of the chapters of my
next book to be published.
Envoi : This skimming through would not be complete
without paper [C72]
regarding the birth
of mathematical morphology, published in 2002, but written in 1998 with G. Matheron and myself. It was his last text, and an answer
to some inaccurate remarks heard at the School of Mines
at that time. As extremes often meet, I added my very first paper on
mathematical morphology, quite in its infancy (1965) [C2].
P.S. the classification marks used
here as links are also those of my publications list, where O, A, C, NI and T
are respectively standing for books, papers, communications with acts,
internal notes and courses. Pdf files that can be
downloaded here are often slightly different from the final publications, as
they consist in first versions, presented in congresses or to journals.
