Title: Super Resolution
Speaker: Uma Mudenagudi
Affiliation: BVB College of Engineering and Technology, Hubli
Date: February 18, 2008
Abstract: We investigate the problem of generating high resolution image or video using information contained in multiple overlapping images or videos of the same scene, whose spatial and temporal resolutions are higher than any of the input low resolution images or videos. This problem is commonly referred to as super resolution, and it provides the possibility of reduction of noise, removal of blur, reduction of motion aliasing, increase of spatial and temporal resolutions. We employ a reconstruction based approach using MRF-MAP formalism, and use approximate optimization using graph cuts to carry out the reconstruction. We give insights on how to keep the super resolution problem well-conditioned. Our results demonstrate that it is possible to obtain super resolution preserving high frequency details well beyond the predicted limits of magnification and reduce one or more degradations.