Feature based image registration using non-degenerate pixels

Image registration (IR) aims to geometrically match one image to another. It is extensivelyused in many imaging applications. Among many existing IR methods, one widely used group ofmethods are feature-based. By a feature-based method, a number of relevant image features arefirst extracted from the two images, respectively, and then a geometric matching transformationis found to best match the two sets of features. However, proper identification and extractionof image features turns out to be a challenging task. Generally speaking, a good image featureextraction method should have the following two properties: (i) the identified image featuresshould provide us proper information to approximate the geometric matching transformationaccurately, and (ii) they should be easy to identify by a computer algorithm so that the entirefeature extraction procedure is computer automatic. In this paper, a new type of image featuresis studied, which has the two properties described above. Together with the widely used thinplate spline (TPS) geometric transformation model, it is shown that our feature-based IR methodworks effectively in various cases.