IEEE
Transactions on Circuits and Systems on Video Technology,
vol. 12, no. 5, pp.331-341, May 2002
Image Information
Restoration Based on Long-Range Correlation
1 Department of Computing, The Hong Kong
Polytechnic University, Hong Kong, China
2 Dept. of Electrical & Computer
Engineering, The University of Texas at Austin, Austin TX, USA
Abstract: A new class of image
information restoration algorithms virtually different from traditional
techniques are proposed. Our methods are originated from the idea that there exists abundant
long-range correlation within natural images and the human vision systems
composed of our eyes and brains can sufficiently utilize such kind of
information redundancy to implement the functions of image interpretation,
representation, restoration, enhancement and error concealment. Our general approach can be summarized as five basic steps:
Fetching, Searching, Matching, Competing and Recovering. The experimental results on several
practical applications show that our methods perform substantially better than
many other state-of-the-art methods.
Index Terms: image processing, image restoration, image enhancement, error
concealment, impulse noise cancellation, blocking effect