Hello:
I am a researcher in Artificial Intelligence currently looking for existing computer programs (or algorithms) able to automatically detect lines or approximate-lines in matrixes; more precisely:
- in the simplest situation, I deal with numerical 2-D (classical) boolean matrixes, i.e. with elements 1 or 0, such as dot matrixes obtained from aminoacid sequence in molecular biology. The problem is to detect line segments typically 5-6 elements-long as on fig. 1 below, or discontinuous/approximate segments as on fig.2:
oooooooooooooooo oooooooooooooooo
ooooo1oooooooooo oo1ooooooooooooo
o1ooooo1oooooooo ooo1oooooooooooo
oo1oooooo1oooooo oooo1ooooooooooo
ooo1ooooooo1oooo oooooooooooooooo
oooo1oooooooo1oo ooooooooo1oooooo
ooooo1ooooooooo1 oooooooooo1ooooo
oooooooooooooooo ooooooooooo1oooo
Fig. 1 Fig.2
- in more complicated ones, I deal with *multidimensional*-, symbolic-elements- matrixes. Here also I am trying to detect alignments (pseudo-linear clusters) of elements having equal attribute values (e.g. elements all having a "color-attribute" value equal to "red" and a "shape-attribute" value equal to "triangle").
A brief literature search in the fields of either computer vision, pattern recognition and molecular biology showed that a number programs/algorithms accomplishing this type of tasks exist, but it's hard to have a clear view so as to focus on the right one(s) for my particular problem.
Anyone's help will be greatly appreciated. Thanks.
-------------------------------
Pierre-Yves ROLLAND
LAFORIA/IBP/Universite Paris 6
4 place Jussieu
75005 Paris, FRANCE
-------------------------------
