A Non-Iterative Registration Method Based On Image Transformation Descriptors

In this article we present a novel method of the image registration based on the image transformation descriptors which are also introduced in this paper. The algorithm computes the registration parameters directly, on the contrary to the most known, iterative methods. It is fast and robust and therefore can be used as an initialization for the more accurate (but slower) algorithms. Our registration technique relies on the values obtained from appropriate transformation descriptions (image transformation descriptors, ITDs), computed from the geometric properties present on the images. It is also possible to adapt the process to a specific registration problem by providing custom descriptors, specific for a particular case. To prove the usability of the method, we have successfully tested it for registering 3D CT images with the angiography of thoracic aortas.