High Frequency Component Compensation based Super-Resolution Algorithm for Face Video Enhancement

This paper proposes a video-based super-resolution algorithm by high-frequency component compensation. Normalized logarithm distance (NLD) minimization on local shape feature is proposed for image registration. Lost high-frequency information are estimated using the registered frames. By compensating the high-frequency component, the high-resolution images are recovered. The algorithm has lower computational cost than the alternatives. Experimental evaluation verified the usefulness of the algorithm.

• Type: Conference
• Year: 2004
• Where: ICPR
• Authors: Junwen Wu, Mohan M. Trivedi, Bhaskar D. Rao
• Keywords: Computer Vision, ICPR 2004, Lower Computational Cost, Video-based Super-resolution Algorithm