comp.lang.idl-pvwave archive: archive » Singular Value Decomposition in 3 Dimensions

Home » Public Forums » archive » Singular Value Decomposition in 3 Dimensions

Show: Today's Messages :: Show Polls :: Message Navigator
E-mail to friend

Re: Singular Value Decomposition in 3 Dimensions [message #62367 is a reply to message #62289]

Wed, 03 September 2008 04:56

Juggernaut
Messages: 83
Registered: June 2008

Member

On Sep 3, 7:52 am, Bennett <juggernau...@gmail.com> wrote:
> On Sep 2, 12:33 pm, tomandwilltam...@gmail.com wrote:
>
>> I am wondering how to do Singular Value Decomposition in 3 Dimensions
>> in IDL. All of the canned routines seem to work only on 2D arrays.
>
>> Specifically, I am trying to preform Principle Component Analysis on
>> stacks of 2D images.
>
>> For example, how can one preform an SVD on a 2048x2048xn array to get
>> 2048x2048 principle components?
>
>> Thanks much,
>> -Will
>
> If you want the principal components for the
> 3D array you can do something like this
> sz = size(array, /dimensions)
> newArray = fltarr(sz[2], sz[1]*sz[0])
> FOR i=0, sz[2]-1 DO BEGIN
> newArray[i,*] = transpose(reform(array[*,*,i], sz[0]*sz[1]))
> ENDFOR
> result = pcomp(newArray, eigenvalues=evals, /standardize)
>
> pcomp() is IDLs built in for doing PCA and result will be
> an array of I believe the same dimensions of newArray which to
> get back into viewing form you could just reform it back like
>
> tv, reform(result[0,*],sz[0],sz[1])
>
> There may be better ways of doing it but I may as well give
> you a point to jump off of

By the way the for loop can be eliminated by just putting
transpose(reform(array, sz[0]*sz[1], sz[2])) into pcomp

> FOR i=0, sz[2]-1 DO BEGIN
> newArray[i,*] = transpose(reform(array[*,*,i], sz[0]*sz[1]))
> ENDFOR
> result = pcomp(newArray, eigenvalues=evals, /standardize)

becomes

result = pcomp(transpose(reform(array, sz[0]*sz[1], sz[2])), ...)

Report message to a moderator

[Message index]

		Singular Value Decomposition in 3 Dimensions By: tomandwilltamu08 on Tue, 02 September 2008 09:33
		Re: Singular Value Decomposition in 3 Dimensions By: Juggernaut on Wed, 03 September 2008 04:56
		Re: Singular Value Decomposition in 3 Dimensions By: Mort Canty on Wed, 03 September 2008 04:46
		Re: Singular Value Decomposition in 3 Dimensions By: Juggernaut on Wed, 03 September 2008 04:52

Previous Topic:	widgets sporadically not created.
Next Topic:	Re: Passing Batch File Names

-=] Back to Top [=-

[ Syndicate this forum (XML) ] [

] [

]

Current Time: Wed Oct 08 19:14:23 PDT 2025

Total time taken to generate the page: 0.00447 seconds