Feature-Based Image Registration

Abstract

Image registration is the fundamental task used to match two or more partially overlapping images taken, for example, at different times, from different sensors, or from different viewpoints and stitch these images into one panoramic image comprising the whole scene. It is a fundamental image processing technique and is very useful in integrating information from different sensors, finding changes in images taken at different times, inferring three-dimensional information from stereo images, and recognizing model-based objects. Some techniques are proposed to find a geometrical transformation that relates the points of an image to their corresponding points of another image. To register two images, the coordinate transformation between a pair of images must be found. In this thesis, a feature-based method is developed to efficiently estimate an eight-parametric projective transformation model between pairs of images.

The proposed approach applies wavelet transform to extract a number of feature points as the basis for registration. Each selected feature point is an edge point whose edge response is the maximum within a neighborhood. During the real matching process, we check each candidate pair in advance to see if it can possibly become a correct matching pair. Due to this checking, many unnecessary calculations involving cross-correlations can be screened in advance. Therefore, the search time for obtaining correct matching pairs is reduced significantly. Finally, based on the set of correctly matched feature point pairs, the transformation between two partially overlapping images can be decided.