Data-driven registration for local deformations

Stefan Huwer

doi:10.1016/0167-8655(96)00052-9

Data-driven registration for local deformations

1996-08-01

6

PRE

AI

DOI：10.1016/0167-8655(96)00052-9

原文链接

原文求助

摘要

En 中文

We present a new data-driven feature-point registration method for image sequences. This method is based on a modified Kohonen network model that performs local non-affine transformations based on a point's neighbourhood context. This new method is ideal for non-real time applications where the situation is too complex to enable an operational modelling of the transformations needed for a point-to-point mapping from one image into another such as in contrast-enhanced sequences of MRI-mammogramms. This new method is compared to existing approaches. The results of tests with different types of grey-value images are discussed.

Keyword:

image registration

data-driven image matching

non-affine transformations

neural networks

deformation matching

Data-driven registration for local deformations

摘要

AI总结

期刊

机构

引用论文