Re: [Teem-users] reading raw data as nrrd

[Gordon L. K. <gl...@uc...>]

Hi-

On Nov 9, 2010, at 5:37 PM, Amit Chourasia wrote:

> Hello,
>
> Essentially the data layout is
>
> Header (fixed bytes = H )
> Data Sample + Binary junk (fixed bytes = n )
> Data Sample + Binary junk (fixed bytes = n )
> Data Sample + Binary junk (fixed bytes = n )
> Data Sample + Binary junk (fixed bytes = n )
> Data Sample + Binary junk (fixed bytes = n )
> ..
> .so on

If the size of "Data Sample" is not fixed, then nrrd is not a lot of  
use; it cannot handle any kind of ragged arrays.  You can use "unu  
make" with a byteskip designed to skip over H + Nn bytes plus the size  
of the first N "Data Samples", in order to get to the N+1th Data Sample.

If the size of "Data Sample" is in fact fixed, call it D bytes long,  
and let ther be N "Data Samples".  Then you can:

unu make -bs H ... -s <D+n> <N> \
  | unu crop -min 0 0 -max <D-1> M \
  | unu reshape -s ...

So you use crop to remove the per-Data-Sample junk (treating each Data  
Segment as a 1D scanline), and then reshape what's left as the array  
you know it to be.

Gordon