A comparison of template matching registration methods for polarimetric imagery

Automatic registration between channels in a passive polarimetric imaging system can be frustrated by naturally occurring intensity variations due to target polarization content. In this paper, three image registration algorithms are tested using polarimetric data in various stages of translational misregistration. These algorithms, known collectively as template matching methods, are cross-correlation, maximum mutual information, and phase correlation. Test data are derived from laboratory and HST Faint Object Camera (FOC) polarimetric imagery. Mutual information registration is shown to be robust against channel intensity inversions at a cost of significantly increased processing time. Conversely, cross-correlation provides the most consistent performance in low signal-to-noise conditions. (1 2).