Image Super-resolution using Multi Image Multi-coupled Dictionary Learning

Hemant S. Goklani, Shravya S., and Jignesh N. Sarvaiya

Keywords

Image super-resolution, Coupled Dictionary Learning (CDL), Cluster specific, Multi Image Multi-Coupled Dictionary Learning (MICDL)

Abstract

Image super-resolution is the process of reconstructing the high resolution image from a low resolution image. This paper proposes Multi Image Multi-Coupled Dictionary Learning (MICDL) as an extension to the algorithm of image super-resolution based on Coupled Dictionary Learning (CDL). In the proposed method, the training lowhigh resolution images are obtained from image database. Along with improving the training stage, the CDL approach is further enhanced in this paper by incorporating clustering. The image patches are clustered using K-means algorithm followed by cluster specific dictionary learning and cluster specific feature space mapping update. Simulation results shows that the proposed MICDL method performs better compared to other existing super-resolution methods.

Important Links:

Go Back