|
From: Paolo C. <pao...@is...> - 2008-04-04 18:31:18
|
Just this morning I find some time to check the status of the various plugins and I was positively impressed by this one :) I have played with the notoriously nasty conrod.stl file (look among the samples) and with a sequence of remove 2manif, reorient, [midpoint subdiv, tri optimiz, planar laplacian]* you can succeed to get a nicely remeshed model. Not bad! Michele wrote: > I recently committed a version of the laplacian smooth that preserves > the shape of the surface. The update of the vertex position is done > only if the average displacement of normals of adjacent faces (from > old to new faces) is less than a given threshold. > > Seemed me ok. I had in mind a slightly different solution (the update of the vertiex position is done only if the new position is not too far from the planes of the original faces), but the resulting functionality is almost the same. I have some doubts on the smoothing of border vertices, that should require a particular management. There is a strange behaviour when operating on selected faces only... > The class for flip planar (PlanarEdgeFlip) now contains an additional > parameter template, which allows you to set the function to calculate > the quality of triangles. > > Seen. Nice :) Quite in the spirit of the library. probably some measure that penalize in some way the number of incident triangles over each vertex should be added. Trying to apply flip, even if they worsen a bit the triangulation quality but they make it more "topologically regular" (in the sense of having almost everywhere vertexes with 6 incident triangles) would be a great add. > Now I am proceeding with the development of mesh improvement by edge > flips based on curvature, trying to avoid the use of VF adjacency. I > modified class CVertexO locally, in order to make available fields H > and K (I added to Curvaturef template parameters) and to test the > changes. For now I am holding the parameter A in the quality field of > vertex, of course it is only a temporary solution, waiting the > availability of fields for the curvature, scheduled for version 1.2 of > meshlab. With some fixing bugs, the new code should work. Obviously, > this change can not be committed without the changes I made locally. > > I have just committed the inclusion of the curvature data in the optional members of the vertex. the adding of the optional data to vertex lead me to discover a NASTY "using namespace vcg"in the meshmodel.h. I wisely removed it triggering one zillion of trivial compilation errors. I patched hopefully all of them. I have already changed the colorize filter to use the new curvature data (that means that the curvature.h files under the meshcolorize dir are no more useful (removed, you can find them in the attic of cvs) Other filters (like the meshsegment) have still to be updated. Best P. -- Paolo Cignoni -- Senior Researcher Visual Computing Laboratory - ISTI - CNR http://vcg.isti.cnr.it/~cignoni ISTI - CNR Via Moruzzi 1, 56124 Pisa ITALY |
