ESTIMATING BAND-TO-BAND MISREGISTRATIONS IN ALIASED IMAGERY

We compare two classes of techniques, cross-covariance-based and Fouri er-based, for estimating band-to-band misregistrations in multispectra l imagery. We show that both methods often give biased estimates of th e misregistrations, the former because of inadequate interpolation pro cedures and the latter because they do not account for the presence of aliasing. Such aliasing is often present, especially in remote sensin g imagery. We describe a Fourier-based method that accounts for aliasi ng and that, for a variety of 512 x 512 image pairs, gives misregistra tion estimates with standard errors quite often less than 1/100th of a pixel in both horizontal and vertical directions. The theory is appli ed to one artificial and three real image pairs, thus demonstrating so me of its practical consequences. There is also a brief discussion of the implications of the theory for image registration. (C) 1994 Academ ic Press, Inc.