This talk describes one view of how and why astronomers resample images in a variety of different contexts. Image resampling is needed on scales spanning the entire sky to micro-arseconds, for well-sampled and highly undersampled images, in the high and low signal to noise regimes, and for continuous and discrete data. In all of these the effects of image geometry can sharply distinguish two-dimensional resampling from the simpler one-dimensional case.
The talk includes an overview of the strengths and weaknesses of some of the commonly used algorithms. Redistribution approaches like drizzle and exact area resampling are discussed in relationship to the standard models for resampling. The talk concludes with a look at how image clipping tools (e.g., the Sutherland-Hodgeman or Liang-Barsky algorithms) may be useful in speeding up and generalizing some sampling algorithms.