Interactive Super-Resolution through Neighbor Embedding

Jian Pu, Junping Zhang, Peihong Guo, Xiaoru Yuan

Abstract: Learning based super-resolution can recover high resolution image with high quality. However, building an interactive learning based super-resolution system for general images is extremely challenging. In this paper, we proposed a novel GPU-based Interactive Super-Resolution system through Neighbor Embedding (ISRNE). Random projection tree (RPtree) with manifold sampling is employed to reduce the number of redundant image patches and balance the node size of the tree. Significant performance improvement is achieved through the incorporation of a refined GPU-based brute force kNN search with a matrix-multiplication-like technique. We demonstrate 200-300 times speedup of our proposed ISRNE system with experiments in both small size and large size images.