Images and image sequences need to be defined on discrete grids to be processed by digital computers. What are the "best" sampling strategies, and how can we efficientely change the sampling structure of an image? In this talk, I will (try to) answer to these and other questions, that arise in fields like image processing, coding, and computer vision. I will begin introducing the foundations of multidimensional sampling theory: basic results on lattice theory, sampling geometry conversion, implementation issues. I will then cover more advanced topics such as: adaptive subsampling of images; spatio-temporal grids for moving images; motion estimation from interlaced sequences; non-conventional sampling structures.