Using multiple orientational filters of steerable pyramid for image registration

Abstract Images acquired by different methodologies are fused to achieve more accurate information and a comprehensive understanding about the objects. With an increasing number of sensors involved, the signal acquisition and analysis system will benefit from the technique named sensor fusion or data fusion. Many successful applications have been found in the areas of computer vision, remote sensing, medical diagnosis, non-destructive testing, etc. Before the image fusion operation, a preprocessing, i.e. registration, needs to be performed, which ensures that all pixels are associated with corresponding physical points of the object. In this paper, we present an automatic two-dimensional image registration method by using a so-called steerable pyramid transform. Based on the steerability of the transform, image features along certain orientations can be obtained. Thereafter, an edge detector, such as Sobel, can further extract oriented edge features of the image. A coarse-to-fine procedure is adopted to utilize these features to achieve the registration result. The effectiveness of the proposed method is illustrated by the experimental results.