Super-Resolution from Multiple Images Having Arbitrary Mutual Motion
Assaf Zomet, Bezalel Peleg · Kluwer Academic Publishers eBooks · 2005
Normal video sequences contain substantial overlap between successive frames, and each region in the scene appears in multiple frames. The super-resolution process creates high-resolution pictures of regions that are sampled in multiple frames, having a higher spatial resolution than the original video frames. In this chapter existing super resolution algorithms are analyzed in the framework of the solution of large sparse linear equations. It is shown that the gradient of the function which is minimized when solving these equations can be computed by means of image operations like warping, convolutions, etc. This analysis paves the way for new algorithms, by using known gradient-based optimization techniques. The gradient is computed efficiently in the image domain instead of multiplying large sparse matrices. This framework allows versatile imaging conditions, including arbitrary motion models, camera blur models, etc. Prior knowledge can also be combined efficiently for obtaining a MAP super resolution solution. As an example, super resolution is implemented with the conjugategradient method, and a considerable speedup in the convergence of the algorithm is achieved compared to other methods.