What is it about?
Intensity-based image registration is successful in image registration, especially when the images are acquired from different sensors. One of their main disadvantages is the computational cost. In this paper, we examine a way to reduce the computational cost using surrogate model based on machine-learning-regression.
Featured Image
Why is it important?
Despite the research in metaheuristic optimization used in image registration, no on has used surrogate models for reducing the computational cost. In our work, we used several methods for regression. Of all of them, support vector regression performed best in constructing a surrogate model of the objective function that is use as a similarity measure for our images. We found that the use of support regression as an approximation of our function reduces the computational cost by 47% without quality loss in our examples.
Perspectives
I hope this article will open new doors regarding the use of surrogate models in intensity-based image registration methods. The reduction of the compatational cost will bring these methods (in terms of duration) closer to the feature-based registration methods, without sacrificing quality for speed.
Constantinos Spanakis
Read the Original
This page is a summary of: Machine Learning Regression in Evolutionary Algorithms and Image Registration, IET Image Processing, February 2019, the Institution of Engineering and Technology (the IET),
DOI: 10.1049/iet-ipr.2018.5389.
You can read the full text:
Contributors
The following have contributed to this page
