Image registration is one of the most fundamental problems in computer vision.
It consists in finding the transformation that aligns two or more images.
Many techniques have been proposed to solve this problem.
We start this chapter by reviewing the state-of-the-art approaches to image registration.
In particular, the two fundamental approaches for parametric image registration, i.e. the feature-based and the direct approaches, will be introduced.
Then, we will present two of our contributions related to image registration.
The first one is a method that allows one to use a direct approach to image registration without needing a region of interest.
This first contribution uses a robust framework for direct image registration based on M-estimators.
It has been published in (37,34,38).
The second one is a new method to automatically tune the hyperparameters that naturally arise when using a feature-based approach to register images.
The novelty of this approach lies in the fact that it uses the pixel information in addition to the features.
It has been published in (35,36).
Note chapter 6 will also deal with image registration.
The main difference of this next chapter is that it will be about defining a parametric image deformation model instead of the parameter and hyperparameter estimation as the current chapter.
Contributions to Parametric Image Registration and 3D Surface Reconstruction (Ph.D. dissertation, November 2010) - Florent Brunet
Webpage generated on July 2011
PDF version (11 Mo)