K-means Clustering을 이용한 효율적인 Learning 기반 Super-Resolution 알고리즘

초록

This paper proposes a computationally efficient learning-based super-resolution algorithm using k-means clustering. Conventional learning-based super-resolution requires a huge training set for reliable performance, which brings out a tremendous memory cost as well as a burdensome matching computation. The proposed algorithm significantly reduces the size of the training set by properly clustering the similar patches in the learning phase in terms of L2-norm distance, with negligible performance degradation.