Video denoising without motion estimation using K-means clustering

Abstract

This study presents a novel dictionary pruning algorithm that found dictionaries of optimized size for a given dataset, without compromising its approximation accuracy and performance. It is achieved by applying KSVD (K-means singular value decomposition) algorithm to patches of dictionary. Proposed method optimized dictionary selection, and with KSVD yielded better video denoising than KSVD with fixed dictionary.