The image registration problem is one that appears in a number of different applications, many of which require solutions to be both fast and highly accurate. Many current registration approaches require the discretization of a continuous search space, which necessitates extra computation time and can result in decreased accuracy. Some of these techniques place limits on types of input data or do not perform complete searches. We introduce an image registration technique that works natively on a continuous search space in the class of affine transformations. It efficiently and systematically traverses the search space and guarantees the best registration according to our similarity metric.
展开▼
