[Insight-users] streamingFilter and filter separation into boundaryand inner faces

[Miller, James V (Research)]

Martin, 

The architecture already support this. All the filters that support
streaming and use boundary conditions already take this into account.

The GenerateInputRequestedRegion() method is used to inform the pipeline
that the input to a filter needs to be larger (if possible) than the 
output region.  This ensures the input region will include an pixels
that are outside the output region but needed to ensure proper neighborhood
calculations (without introducing artificial boundary conditions).

The face calculator will divide a region into how every many pieces
are necessary to handle boundary conditions.  The face calculator 
could just return one region, where no pixels have boundary conditions, 
or could return a set of regions, one with not boundary conditions, and the 
rest with boundary conditions.

Jim



-----Original Message-----
From: insight-users-bounces+millerjv=crd.ge.com at itk.org
[mailto:insight-users-bounces+millerjv=crd.ge.com at itk.org]On Behalf Of
Martin Urschler
Sent: Thursday, November 03, 2005 5:44 AM
To: insight-users at itk.org
Subject: [Insight-users] streamingFilter and filter separation into
boundaryand inner faces


hi,

i'm currently looking through the streaming filters and the boundary 
calculator and i'm wondering if it is possible to simultaneously use 
these two concepts in a filter?
i'd like to process my large data sets in chunks on the one hand, 
otherwise i'm getting problems with memory consumption of the 
intermediate results
on the other hand i have convolution filters involved in my streamed 
mini pipeline. so i'd also like to make the distinction between my inner 
face and the boundary faces for efficiency

now i'm not quite sure if and how this works, since the boundary face 
calculator of the streamed chunks should distinguish between boundary 
faces in the interior of the whole large data set and faces on the 
exterior of the whole large data set
otherwise i would get wrong calculations on the interior chunk 
boundaries, since the pipeline would take the boundary value instead of 
the data from the neighboring chunk

is there a mechanism that already handles this issue? is there even a 
filter that already works according to this principle, or presents a 
proper workaround by implementing the right padding (voxel padding in 
this case i suppose, not region padding) in the interior of the volume?

thanks,
Martin

_______________________________________________
Insight-users mailing list
Insight-users at itk.org
http://www.itk.org/mailman/listinfo/insight-users