| Coding for speed help needed [message #9076] |
Wed, 28 May 1997 00:00 |
pford
Messages: 33
Registered: September 1996
|
Member |
|
|
I have been requested to explain what I am try to do and why.
I am using a genetic algorithm to do a tomographic reconstruction of
radionuclide data using a front or forward projection instead of the usual
backprojection method. I am comparing the sinograms generated by the
genetic algorithms "object image" compared to the "target's" sinogram to
determine the 'fitness.' I am doing this as a proof of concept to
calculated
unknown parameters, i.e. attenuation. The B array of bytes is the genetic
code that needs to be decoded. Its size is:
pop_size = 36; this can vary greatly
dim = 64
B = bytarr( dim * dim * 5 * pop_size)
There are pop_size individuals
The first 64*64 (dim*dim) bytes code for the activity or intensity in a 64
by 64 image. The next 4*64*64 bytes code for the attenuation coefficients
which are
between 0.0 and 1.0, where 1.0 is no attenuation and 0.0 is 100%.
I need to strip off the appropriate bytes and convert them.
For example let:
activity_code = bytarr(dim*dim*4)
activity_code(*) = B(dim:dim*dim*4-1); should give me the bytes I want to
;decode to float.
I need something that is fast and consistent. Since it is a genetic
algorithm, the sequencing of the bytes, i.e, which byte is high order vs
lower order, or even if they are not contigious is not important, as long
as it is consistent.
The number of bytes that code for a float is arbitrary, it just affects
the precision. (I am sure I won't need this much precision, but I want to
debug it before fiddling with other parameters.)
In case you are curious, the forward-projection without attenuation does
work.
Thanks for the help so far. I will give it a try.
Patrick Ford, MD
Department of Radiology
Baylor College of Medicine
|
|
|
|
| Re: Coding for speed help needed [message #9077 is a reply to message #9076] |
Wed, 28 May 1997 00:00  |
jackel[1]
Messages: 11
Registered: May 1997
|
Junior Member |
|
|
In article <5mhn9v$r3r@gazette.bcm.tmc.edu> pford@bcm.tmc.edu (Patrick V. Ford) writes:
> What I would like to do, is do this in IDL where the result would go into A,
> where
> A = fltarr(64,64)
> and the byte array is
> B = bytarr(64*64*4)
Not sure if I understand properly, but if your array "B" was of the
form B=bytarr(4,64,64), then doing A= long(b,0,64,64) would turn
each 4 byte chunk into a long word. Check the documentation for
the "offset" argument to "long" (and float and fix etc). If that's
what you want, then try using "transpose" to get "B" into the proper
shape.
Brian Jackel
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
