Performance analysis on multi-frame image Super-Resolution via sparse representation

This paper proposes quality analysis of multi-frame Super-Resolution. We compare three algorithms of multi-frame Super-Resolution such as Bilateral Total Variation, Dual-Dictionary, and Kernel based Principal Component Analysis (KPCA). This research focuses on solving the problem in difference texture images. We experiment on Baboon, Lena, Eye, and Access Road. The algorithms are applied on 16 frames interval at 100 iterations. The experimental results show Peak Signal to Noise Ratio (PSNR) versus the number of iterations. The Bilateral Super-Resolution has the lowest number of iterations with high PSNR in low texture images. The experimental results also show that PSNR drops in Kernel Principal Component Analysis approach. In addition, we have found that the blurring process is an ill posed condition for low texture images.