Fast and Effective Geometric K-Nearest Neighbors Multi-frame Super-Resolution

Multi-frame super-resolution is possible when there is motion and non-redundant information from a sequence of low-resolution input images. Remote sensors, surveillance videos and modern mobile phones are examples of devices able to easily gather multiple images of a same scene. However, combining a large number of frames into a higher resolution image may not be computationally feasible by complex super-resolution techniques. We discuss herein a set of simple and effective high-performance algorithms to fastly super-resolve several low-resolution images in an always-on low-power environment, with possible applications in mobile computing, forensics, and biometrics. The algorithms rely on geometric k-nearest neighbors to decide which information to consider in each high-resolution pixel, have a low memory footprint and run in linear time as we increase the number of low-resolution input images. Finally, we suggest a minimum number of input images for multi-frame super-resolution, considering that we expect a good response as fast as possible.