MATLAB Answers

RNTL
1

extract 3D surface from 3D volume data

Asked by RNTL
on 2 Feb 2012
Latest activity Commented on by GS89
on 3 Apr 2017
Hi, I have a volumetric data (in the form of (x,y,z,val(x,y,z)) array of data on and within a 3D volume shape. I would like to take only the coordinates and value on the surface.
Does anyone have an idea ?

  0 Comments

Sign in to comment.

3 Answers

Answer by Sean de Wolski
on 2 Feb 2012

It sounds like you may want to create an isosurface (a volumetric contour map).
doc isosurface

  4 Comments

Show 1 older comment
RNTL
on 3 Feb 2012
thinking about it, will not work, since the vector from the [0,0,0] always passes through atleast 2 surface points. :(
Now we can start working!
If you don't take the radius from [0,0,0] but from the center of your points [xC,yC,zC] then you might be set. If you should calculate the center points with the mean or median of the points, like the center of mass (sum(x.*val)/sum(val)), also it is not given that such center even are inside a general 3-D body. But these steps you can judge better yourself.
HTH
RNTL
on 10 Feb 2012
Hi,
unfortunately its more complicated than that. I could use it if the shape was surrounded by a closed curve (in 2D plane). but its not the case.

Sign in to comment.


Answer by Bjorn Gustavsson on 2 Feb 2012

Algorithm in steps:
  1. Extract the surface elements in x y z and val
Done! Since we cant know how you define the surface, the shape of the surface or any other facts about it except that it is a surface in 3-D it is not easy for us to give you more than that. If you want to display it you might get somewhere with:
T = delaunay3(X,Y,Z,{'Qt', 'Qbb', 'Qc', 'Qz'} ); and then something like tetramesh
HTH

  3 Comments

RNTL
on 2 Feb 2012
Well step (1.) pretty much is the issue.
I want to separate the surface points from the rest of the volume, the space is the surface and its interior volume.
Yes, and we still do not know how you see the surface, is it the boundary between where val is finite and nan, 0? Something else? If you want us to be able to give even suggestions you'd be better off if you gave us something to go on...
I have the same question as the asker. To answer Bjorn's question, Suppose I pick any threshold, say 0.5, how can I find the surface where the volume = 0.5?

Sign in to comment.


Answer by Kevin Moerman on 10 Feb 2012

Could you rephrase/clarify your question? It sounds like you have 3D image data and that you are only interested in a certain surface within this volume.
Do you have intensity data x,y,z,val for a cloud of points or is that an image? And do I understand it correctly that you have coordinates that you know lie on the surface? Like x_s,y_x,z_s? And that you want to find the intensities on the surface points e.g. val_s? This can be achieved through interpolation (e.g. griddata3 or TriScatterInterp). If you want an actual description of the surface like patch data (faces and vertices) for plotting of the possibly closed shape (you don’t need this for determination of val_s on these points) then its more complex. Then you have point cloud onto which you want to draw a surface. You could try the cart2sph approach you suggested but this will produce possible artefacts at the poles and will lead to non-unique solutions at locations with double solutions in the spherical coordinate system (multiple radii solutions for a given angular coordinate set). There’s a number of quick and dirty approaches to getting the surface: 1) Just use the convex hull 2) Do a Delaunay tessellation (the outer surface elements now describe the convex hull again) but delete tetrahedrons which contain edged which are too long (i.e. these might be bridging the gaps that you do not want bridged). Then find the outer triangles to find your surface. This only works if you have a nice even point spacing. Good luck.
Kevin

  2 Comments

RNTL
on 10 Feb 2012
Hi Kevin,
The prob description is this: I have a set of points which is given according to (x,y,z,intensity), where the (x,y,z) values are their spatial location in cartesian coord system. and the intensity at this point. The data I have should contain solely the on-body and in-body points. however, I would like to extract only those points which are on-body.
1. don't have prior knowledge on the on-body intensities.
2. the shape of the body is not convex, so as i understand the convex hull will not be useful here.
I have to check griddata3, maybe it can help.
Thanks, Ron
GS89
on 3 Apr 2017
Did anyone figure out how this can be done?

Sign in to comment.