Overcoming registration uncertainty in image super-resolution: maximize or marginalize?

Journal article

In multiple-image super-resolution, a high-resolution image is estimated from a number of lower-resolution images. This usually involves computing the parameters of a generative imaging model (such as geometric and photometric registration, and blur) and obtaining a MAP estimate by minimizing a cost function including an appropriate prior. Two alternative approaches are examined. First, both registrations and the super-resolution image are found simultaneously using a joint MAP optimization. Second, we perform Bayesian integration over the unknown image registration parameters, deriving a cost function whose only variables of interest are the pixel values of the super-resolution image. We also introduce a scheme to learn the parameters of the image prior as part of the super-resolution algorithm. We show examples on a number of real sequences including multiple stills, digital video, and DVDs of movies.

Authors: Pickup, L; Capel, D; Roberts, S; Zisserman, A

• Publication status: Published
• Peer review status: Peer reviewed
• Publisher: Springer
• Journal: EURASIP Journal on Advances in Signal Processing
• Volume: 2007
• Article number: 023565
• Publication date: 2007-12-01
• Acceptance date: 2007-05-04
• DOI: 10.1155/2007/23565
• EISSN: 1687-6180
• ISSN: 1687-6172
• Language: English
• Keywords: digital video; imaging model; image registration; cost function; generative imaging