Thread Subject: How to make isosurface() work with uint8 data

I found isosurface() and isonormal() are very useful for
processing 3D data sets. However, my version (2007b), these
functions only take "double" as input, which requires 8
times more storage place.

Does anyone help me how to make isosurface() work with 8-bit
data, so I can process the 3d data in higher resolution?

"Salmon " <yourfellow@yahoo.com> wrote in message
<g02f0e$k84$1@fred.mathworks.com>...
> I found isosurface() and isonormal() are very useful for
> processing 3D data sets. However, my version (2007b), these
> functions only take "double" as input, which requires 8
> times more storage place.
>
> Does anyone help me how to make isosurface() work with 8-bit
> data, so I can process the 3d data in higher resolution?
Dirty and simple? instead of using isosurface() use
isosurface(double()).

A more elegant solution is to create a directory @uint8 in a
directory already in the PATH, and there store a
isosurface.m file which reads like this:

function handles=isosurface(varargin)
%converts any numeric parameter to type double
V=varargin;
for i=1:length(varargin)
    if isa(varargin{i},'extended')
        V{i}=double(varargin{i});
    end
end
handles=isosurface(V{:});
Hope this helps
Regards
Carlos

Oops! replace the word "extended" by "uint8"
Regards
Carlos

Carlos,

Your solution is not the solution I am looking for. It did
not improve any performance. What I meant is that "double"
is 64 bit, while uint8 is 8 bit. You see, there is 8 times
memory and calculation savings. That's very significant
memory saving and speed improvement. Besides, almost all
real-world 3D data sets are either 8 bit or 16 bit.

Is there somewhere that I can find the isosurface source
code, and replace all the double to uint8? I think MW
should make all those 3D related function take 8 bit data.

Thanks,

> A more elegant solution is to create a directory @uint8 in a
> directory already in the PATH, and there store a
> isosurface.m file which reads like this:
> > handles=isosurface(V{:});
> Hope this helps
> Regards
> Carlos
>

BTW: if you look at the isosurface, inside you find that it
actually calls for a mex function isosurf(), which I assume
is a C/C++ or Fortran function. Does anyone know how to get
the source code for that function?

"carlos lopez" <clv2clv_00000000_@adinet.com.uy> wrote in
message <g02j1s$cop$1@fred.mathworks.com>...