| find rotation between two images [message #40702] |
Wed, 01 September 2004 14:54  |
tbeetz
Messages: 9
Registered: May 2003
|
Junior Member |
|
|
i have several images where i would like to find the rotation and
shift compared to a reference image. i use an alignment routine to
seperate the shifting part. however, there is still the remaining
problem of image rotation. is there a routine available? thanks!
|
|
|
|
|
|
|
|
| Re: find rotation between two images [message #40892 is a reply to message #40702] |
Fri, 10 September 2004 14:35  |
dcw_yip
Messages: 22
Registered: October 2003
|
Junior Member |
|
|
Do some simple feature extraction and then do a least-squares fit.
Look up the paper in IEEE transactions entitled "Least-Squares Fitting
of Two 3-D Point Sets" by Arun, Huang and Blostein. I wrote some code
to do basically what you are describing. Their paper is very clear
and took me about an hour or two to code up. You won't get exact
rotation angles. They'll be within a half a degree or so.
There are also some enhancements to their algorithm to give you
scaling, shearing, etc, etc if you need that.
David
tbeetz@yahoo.com (Tobi) wrote in message news:<7f0cc27.0409011354.45cce42f@posting.google.com>...
> i have several images where i would like to find the rotation and
> shift compared to a reference image. i use an alignment routine to
> seperate the shifting part. however, there is still the remaining
> problem of image rotation. is there a routine available? thanks!
|
|
|
|
|
|
| Re: find rotation between two images [message #40898 is a reply to message #40702] |
Fri, 10 September 2004 02:02 |
helaha
Messages: 26
Registered: March 2004
|
Junior Member |
|
|
tbeetz@yahoo.com (Tobi) wrote in message news:<7f0cc27.0409011354.45cce42f@posting.google.com>...
> i have several images where i would like to find the rotation and
> shift compared to a reference image. i use an alignment routine to
> seperate the shifting part. however, there is still the remaining
> problem of image rotation. is there a routine available? thanks!
I have developed an image processing tool IQM and there is an
alignment feature "Align stack". The first image of an image stack is
centered in the middle of the image (simply by shifting the centroid).
The subsequent images are aligned by three steps of shifts and
rotations (each step consists of up to 100 shifts and 100 rotations).
For each step an actual shift and rotation, which minimises the
statistical error is chosen. For large images, the calculation times
are therefore quite long.
If you are interested, look at
http://www.uni-graz.at/people/ahammer/IQM/index.htm
Regards,
Helmut Ahammer
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
