Image Registration

[Yili Zhao]

Pablo wrote:
>The long term goal is to introduce an automatic image registration
>algorithm, either on the back of PanoTools (feature based) or maybe a
>frequency domain approach.

  What do you mean "feature based"? 
  It seems that there are some categories of image registration methods:
  o intensity based or gradient based method (for local optimze)
  o feature based method
  o correlation based method (for global optimize)
  It seems that pt's registration method is to minimize control point distances,
and the distances are measured in the final panorama's space.

   
Edward wrote:
>I think the algorithm is fairly clear though (but not trivial).  
>Basically do rough image matching using a frequency based >alignment method (
>I'm assuming it will be accurate enough given the distortions involved mind...).  
>Then several papers >have referenced using a certain filter (forget the name) 
>to locate hundreds of features in each image.  Match and optimise >these between 
>images (pruning the high error matches since these are likely mis registrations between each image).

May you list the papers' name you have searched for reference?

Some papers I searched:
1  A Survey of Image Registration Techniques.
   A good introduction.
2. QuickTime VR - an image-based approach to virtual environment navigation.
   This paper describes QTVR framework, and says QTVR uses correlation based
method to do image registration, but does not have algorithm's detail.
3. Video mosaics for virtual enviromnets
   This paper introduces intensity based image registration.
4. Construction and refinement of panoramic mosaics with global and local alignment.
   The authors have several papers on this topic.
5. Creating full view panoramic image mosaics and texture-mapped models.
   The same authors like above.
6. Registration, Calibration and Blending in Creating High Quality Panoramas.
   By Ken Turkowski from Apple Inc.

Yili