comp.lang.idl-pvwave archive: archive

Home » Public Forums » archive » Image warping in IDL

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

Image warping in IDL [message #51188]

Wed, 08 November 2006 04:53

Wox
Messages: 184
Registered: August 2006

Senior Member

Hello,

I have a question concerning image warping: "Is there a fast way of
doing forward mapping in IDL?"

To make myself clear, some remarks:

1. IDL's WARP_TRI uses inverse mapping. It uses triangulation and
surface interpolation to get a "non-integer" position in the original
image for each "integer" position, ie. pixel, in the output
(destination) image. The reverse mapping now involves surface
interpolation of the original image at the non-integer positions.

2. The triangulation and surface interpolation step is already done,
so that leaves only the mapping. However the non-integer positions in
the "output" image were calculated in the triang-interpol step for
each integer position, i.e. pixel, in the "original" image.
Additionally, this "triang-interpol" step uses other techniques, using
external information (spline coefficients) that can't be changed.

3. Because of step 2, forward mapping has to be performed.

Possible answers to the question above (+ why there's a problem):

1. Just do the forward mapping the hard way:

; img: original 2D array
;loop over rows
for i=0,nrow-1 do $
resamplearr, xmap[*,i], img, interimg, ncol, 1, i*ncol

; loop over columns
for i=0,ncol-1 do $
resamplearr, ymap[i,*], interimg, img, nrow, ncol, i
; img: destination array

So this loops over all rows, resampeling them separate, creating an
intermediate image. Then it does the same for all columns of the
intermediate. For 2000x2000 arrays, you can imagine how slow it is.

2. Another possibility would be, when we have the xmap and ymap of the
forward mapping, i.e. pixel [0,0] mapped to [xmap[0,0],ymap[0,0]]
...etc., to convert them for inverse mapping, i.e.
[xmap[0,0],ymap[0,0]] mapped to [0,0] ...etc.

The problem is how? You could do this:
img = WARP_TRI( xmap, ymap, indgen(ncol)#replicate(1b,nrow),
replicate(1b,ncol)#indgen(nrow) , img)

But here the number of controlpoints equals the number of pixels in
img. Except for the speed, the general idea seems strange. You use the
"triang-interpol" step on the "triang-interpol", if you know what I
mean.

3. Maybe there is a way of making solution 2 faster. In remark 2, I
stated that the "triang-interpol" is already done using external
parameters. I can't change the parameters (I'm not calculating them)
but maybe I could convert these parameters so "inverse" mapping xmap
and ymap are calculated. The problem is, I wouldn't know how. To be
specific, these parameters are coefficients of 2D splines, one for x
and one for y.

This text is getting longer and longer. My apologies for this. At this
point I would have to thank you for reading this in the first place.
Thanks!

So the initial question becomes now:
1. Is there a way of converting forward mapping splines to inverse
splines?
2. If not, is there a fast way of doing forward mapping (cfr. solution
1)?

I hope this is making any sense. And thanks again.

Report message to a moderator

[Message index]

		Image warping in IDL By: Wox on Wed, 08 November 2006 04:53
		Re: Image warping in IDL By: David Fanning on Fri, 10 November 2006 15:11
		Re: Image warping in IDL By: JD Smith on Fri, 10 November 2006 14:51
		Re: Image warping in IDL By: Wox on Tue, 14 November 2006 02:20
		Re: Image warping in IDL By: JD Smith on Mon, 20 November 2006 12:17
		Re: Image warping in IDL By: James Kuyper on Fri, 10 November 2006 08:38
		Re: Image warping in IDL By: Wox on Fri, 10 November 2006 01:12
		Re: Image warping in IDL By: Wox on Fri, 10 November 2006 01:05
		Re: Image warping in IDL By: JD Smith on Thu, 09 November 2006 11:50
		Re: Image warping in IDL By: Wox on Wed, 15 November 2006 04:18
		Re: Image warping in IDL By: Wox on Tue, 14 November 2006 08:20
		Re: Image warping in IDL By: Wox on Tue, 21 November 2006 01:18
		Re: Image warping in IDL By: JD Smith on Tue, 21 November 2006 09:10
		Re: Image warping in IDL By: Jeff Hester on Mon, 20 November 2006 09:00
		Re: Image warping in IDL By: Wox on Mon, 20 November 2006 01:18
		Re: Image warping in IDL By: JD Smith on Wed, 22 November 2006 08:22
		Re: Image warping in IDL By: Wox on Wed, 22 November 2006 00:21

Previous Topic:	Making movies
Next Topic:	Re: Pair Counts in an Annulus, for large data sets

-=] Back to Top [=-

[ Syndicate this forum (XML) ] [

] [

]

Current Time: Wed Oct 08 14:32:27 PDT 2025

Total time taken to generate the page: 0.00450 seconds