Super-Resolution Enhancement of Video

Christopher M. Bishop

 

Microsoft Research
7 J J Thomson Avenue
Cambridge, CB3 0FB, U.K.
cmbishop@microsoft.com
http://research.microsoft.com/~cmbishop

Andrew Blake

 

Microsoft Research
7 J J Thomson Avenue
Cambridge, CB3 0FB, U.K.
ablake@microsoft.com
http://research.microsoft.com/~ablake

Bhaskara Marthi

 

University of California, Berkeley

bhaskara@cs.berkeley.edu

http://www.cs.berkeley.edu/~bhaskara

 

 

 

 

 




Published in Proceedings Ninth International Conference on Artificial Intelligence and Statistics,
Key West, Florida, January 3-6, 2003.

Sample videos

These are the three example video sequences used as input to the algorithms.

Example 1
Example 2
Example 3

Algorithm A

This algorithm simply restores each frame independently as discussed in Section 2.4 of the paper. Severe flickering can be observed in both the foreground and the background.

Example 1
Example 2
Example 3

Algorithm B

This algorithm uses the stasis prior, described in Section 3 of the paper. The flickering in the background has been significantly reduced.

Example 1
Example 2
Example 3

Algorithm C

This algorithm uses the motion prior, described in Section 4 of the paper. There is a reduction in foreground flicker at the expense of some return of flicker to the background.

Example 1
Example 2
Example 3