Registration for Super-Resolution: Theory, Algorithms, and Applications in Image and Mobile Video Enhancement
In this chapter, the authors present subpixel image registration that is needed in order to be able to correctly reconstruct any high resolution information. They discuss their camera model, and how super-resolution can be applied to images captured with such cameras, followed by a definition of what we understand by the term “resolution.” The authors present super-resolution as a multichannel sampling problem with unknown offsets. The relation between sensor resolution and the optics of a digital camera is determined by the Nyquist sampling theorem: the sampling frequency should be larger than twice the maximum frequency of the image content coming out of the optical system. Super-resolution algorithms typically combine multiple aliased images with small relative motion, and create a single high-resolution image. The set of images used in a super-resolution algorithm can also be a video sequence, where the motion between subsequent frames is typically small.