Image registration is used to bring objects visible in two or more images into spatial correspondence. Here, one of the images serves as a reference for finding the geometric transformation of objects in the other image(s). Images can then be resampled such that objects positions get constant on all the images.

Medical image registration is for most cases not optimally solved. We tend to improve registration approaches by developing algorithms that enable usage of prior knowledge specific for the task. Our research focuses mainly on advanced methods of local estimation of criterion functions (image similarity) for multi-modality cases, based on point similarity measures.

A part of this project is devoted to establishing a development environment in a form of Matlab/Octave toolbox that is easy to use and understand and includes all the required functionality to build and adopt registration methods for specific tasks.
With a focus to image registration, we participate in COST Parenchima action CA 16103, working group 2, whose goal is to deliver an R&D Toolbox consisting of a coherent set of databases and software needed for further development of renal MRI biomarkers.

 

REG-toolbox Github: https://github.com/progelj/REG-toolbox
Projekt vodi: doc.dr. Peter Rogelj