[Insight-users] MPI Vector Gradient Anisotropic Diffusion Image Filter

[Kevin H. Hobbs]

On 12/06/2010 02:25 PM, Bradley Lowekamp wrote:
> Hello Kevin,
> 
> The warning is related to the pixel type you are streaming/pasting to
> the output file vs the pixel type in the file. MetaImageIO reports all
> multi-component images as images of vector pixel type. RGBPixels are not
> vector pixel types. So this warning is saying you are streaming
> RGBPixels to a Vector image, which doesn't match.  If you changed the
> output image to a image of vector is should go away. 
> 

I _think_ I understand.

> I think this is reasonable behavior and a reasonable warning for this
> situation. What do you think?

I'm going to try to simply stream read/write an RGB image if that gives
the warning for each piece then I think something is unreasonable.

> 
> BTW: I am curious of you are communicating between each node after each
> iteration, or are just having a large overlapping region?

I am communicating the sum, pixel count (silly this dosn't change), and
average of the gradient magnitude of each unpadded piece after each
iteration.

Each node works on a region which is bigger than it's output region by
the number of iterations in each direction.

I suppose I could save some RAM and computation time by communicating
pixel values in a 1 pixel wide band but that seemed like too much
hassle. (Who wins at the corners?) I tried this for MPI Fast Marching
and it was a nightmare even though I had an algorithm to follow.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 253 bytes
Desc: OpenPGP digital signature
URL: <http://www.itk.org/pipermail/insight-users/attachments/20101206/4847abf5/attachment.pgp>