Resampling considerations for registering remotely sensed images

Satellite imagery data typically requires the use of interpolation techniques during the orthorectification and registration stages of processing. Given that satellite imagery is typically undersampled, interpolation cannot perfectly recover the original signal and therefore consideration must be given to the concept of the optimal interpolation technique for a given application. We address these issues and suggest a metric to define interpolation quality. We also propose a new interpolation scheme which performs well for certain classes of image features, most importantly sub-pixel image features and features of low pixel count.