Re: Subpixelshift by Phase correlation
- From: serg271 <serg271@xxxxxxxxx>
- Date: 25 Apr 2007 04:27:48 -0700
On Apr 24, 1:18 pm, "jalexn...@xxxxxxxxxxxxxx"
<jalexn...@xxxxxxxxxxxxxx> wrote:
Hi all
I am presently working on subpixel shift detection using phase
correlation .The algorithm uses the shifting property of the FT , to
find the shifts between the images .
Based on the algorithm the IFFT( phase correlation), should yield a
delta function at the displacement required to register the two
images.
Problem:
I am dividing the images into subblocks and estimating the shift
individually, Now instead of single delta function (peak) at the
displacement required to register the two images , I do get a sharp
dominant peak at (0,0) frequency and other random peaks.
This problem occur only with real time images, with simulated images
the dominant peak at (0, 0) is suppressed and the algorithm works
correctly.
Can anybody throw some light on the why there is a dominant peak at
the (0,0) frequency.
You should get a dominant peak at the shift frequency and some
secondary peaks around it.
If you get dominant peak in the wrong position it mean either you have
some bug in the algorithm, or you images are too bad - phase shift is
too big relatively to image size, or noise too strong.
.
- Follow-Ups:
- Re: Subpixelshift by Phase correlation
- From: jalexnoel@xxxxxxxxxxxxxx
- Re: Subpixelshift by Phase correlation
- References:
- Subpixelshift by Phase correlation
- From: jalexnoel@xxxxxxxxxxxxxx
- Subpixelshift by Phase correlation
- Prev by Date: Re: Corner features in pictures using Harris detector
- Next by Date: Re: Corner features in pictures using Harris detector
- Previous by thread: Re: Subpixelshift by Phase correlation
- Next by thread: Re: Subpixelshift by Phase correlation
- Index(es):
Relevant Pages
|
