MATLAB Central Newsreader - image processing

image processing

Peter Schreiber — Sat, 07 Feb 2009 17:15:04 -0500

Hi guys,
Is there any way to load pictures (any format would be okay, but high resolution) into matlab and rotate them in a 3d cartesian coordinate system? I have to make a 3d picutere, consisting of slices at different focal positions, where on each slice the intensity of the beam is shown.

Best Regards,
Peter

Re: image processing

ImageAnalyst — Sat, 07 Feb 2009 22:48:48 -0500

On Feb 7, 12:15=A0pm, "Peter Schreiber" <schreiber.pete...@gmail.com>
wrote:
> Hi guys,
> Is there any way to load pictures (any format would be okay, but high res=
olution) into matlab and rotate them in a 3d cartesian coordinate system? I=
have to make a 3d picutere, consisting of slices at different focal positi=
ons, =A0where on each slice the intensity of the beam is shown.
>
> Best Regards,
> Peter

------------------------------------------------------------------
Peter:
imrotate only rotates in 2D, so that won't help. imtransform() could
probably give you a 2D image of a projection of an image rotated in 3D
(in which case the 2D projection would look like it has a perspective
(vanishing point) in addition to the rotation.) You might try
tformarray() which can apply a transform to an N-D array. You may
have to load your 2D pictures into a 3D volume image array first
though (use the cat() function). I can't help you beyond that, as I
haven't used tformarray, so just try to understand the help and
examples for it. These all require the image processing toolbox.
Good luck,
ImageAnalyst

Re: image processing

Pete — Sun, 08 Feb 2009 19:06:02 -0500

If I understand you correctly, one method would be to use SURF to display planes and use your image data to provide the colours/gray levels. For example, if your image is in a variable IM:

offset = 0;
[xx, yy] = meshgrid(1:size(im,2), 1:size(im,1));
zz = ones(size(xx)) * offset;
surf(xx, yy, zz, im, 'EdgeColor', 'none');
colormap gray

You can have multiple planes by varying the offset. You can also control the transparency of each plane with the 'FaceAlpha' property. The interactive controls will allow you to rotate in 3D, although I suspect that would be very slow for high resolution images. It could help to determine a nice display angle on a very low resolution version of your images, store it using the VIEW function and then use VIEW again to set the angle directly from the command line for the final plot.

Re: image processing

Image Analyst — Sun, 08 Feb 2009 20:29:02 -0500

Pete
That wasn't what I was thinking. but who knows it might work for surfaces. Well for us, we don't bother with MATLAB for true 3D volume visualization. It's just not made for such things, as compared to others that are, such as
avizo: http://www.tgs.com/products/avizo.asp
amira: http://www.amiravis.com/
volvis: http://www.cs.sunysb.edu/~vislab/volvis_home.html
volview: http://www.kitware.com/products/volview.html
With these you can really do powerful volume visualization including cut-away views, projections, segmentation, transparent rendering, threshold-based rendering, rotation (like you want), cropping, resizing, converting to other formats, distance and volume measurement, and so much more.

If you really want to use MATLAB instead of some of the superior packages, then Google has quite a few ideas ("volume visualization in MATLAB" was one of it's automatically generated hints/suggestions):
http://www.google.com/search?hl=en&q=volume+visualization+in+matlab&aq=5&oq=volume+vis

Regards,
ImageAnalyst