# comp.graphics.algorithms

## Subject: Re: Mesh Correction Algorithm - Academic Question

Check out Greg's papers:
http://www-static.cc.gatech.edu/~turk/pubs.html

In particular:
http://www-static.cc.gatech.edu/~turk/my_papers/retile.pdf

- Saty

> Hi all,
>
> about the following problem that I face as part of my research.
>
> I am running an algorithm (which I can't really talk too much about)
> which achieves the following. Given a set of lets say point cloud data
> (points which are known to lie on a surface), my algorithm takes a
> random triangular mesh and fits it to this data, such that on conclusion
> of the algorithm the mesh represents the surface from which the range
> data was taken. Now my algorithm achieves what I will describe as a good
> representation of the surface.
>
> However, two problems seem to arise,
>
> 1) In small areas, some faces of the mesh are such that the following
> sort of thing has occurred. The vertices of these faces appear to
> reasonable (i.e. sensibly position for the surface), however the faces
> have what can best be described as collapsed ontop of each
> other(Difficult to describe in text). So two faces lie across each
> other. Now I could simply take the final vertex positions of all faces
> and show how retriangulate the surface, but the vast majority of faces
> are fine.
>
> 2) The concentration of vertices can be highly bunched in certain
> localised areas, i.e. non even concentration across the face. I am not
> looking for an even distribution, but in some areas that are fairly
> planar they are really way too many vertices (and hence faces) which I
> think is redundant information. Obviously areas of great variation in
> surface need to be adequately represented.
>
> Could anybody suggest a solution, for 1) some sort of retriangulate
> certain problematic areas and 2) some sort of smoothing, to attempt give
> a better overall surface likeness.
>
> Visually the problem been at the moment, when I texture my final mesh it
> can sometimes appear a little bumpy.
>
> For 2) I have tried the Jacobi Laplace Smoother available through
> OpenMesh (I know nothing really about mesh smoothing), but found that
> this technique redistributed my vertices but also moved them such they
> no longer represented the surface particularly well.
>
> Now I understand this is hard to imagine in text alone and anybody
> wanting to help I could send a mesh or picture if that made things clearer.
>
> Any help would be much appreciated. I doubt anybody has the magic bullet
> to this problem but a discussion / starting suggestions is what I really
> need at this moment.
>