Multimedia Error Concealment by Projection onto Convex Sets (POCS) based Methods

Andrew Cheng1 and Rohit Puri2
(Professor Kannan Ramchandran)
Intel

In this work we present an error concealment scheme for damaged spatial blocks in image/video sequences. A promising error concealment approach in such a scenario is based on the projection onto convex sets (POCS) [1] method. If the two convex sets intersect, this algorithm guarantees a reconstructed block that will be smoothly connected with adjacent blocks without altering the correctly received adjacent blocks. If the two sets don't intersect, the algorithm oscillates between two points that are close to both properties. The POCS method has the additional benefit of edge continuity and preservation via edge detection and adaptive filters. However, one shortcoming of the algorithm is that it assumes a damaged block is surrounded by undamaged blocks. In this work we deal with the case where damaged blocks often are contiguous or located on the boundary of the image. This has potential application in some recently proposed low-complexity video coding algorithms [2,3] that rely on the coding with side information approach [4] to achieve compression as opposed to conventional motion compensation based approaches. We seek to generalize the POCS algorithm to incorporate these cases. Furthermore, we are investigating various methods of providing an initial estimate of the missing block for the POCS algorithm. Simple averaging techniques have been remarkably successful for individual missing blocks, but more elaborate methods are needed as the number of contiguous missing blocks increases.

[1]
H. Sun and W. Kwok, “Concealment of Damaged Block Transform Coded Images Using Projections onto Convex Sets,” IEEE Trans. Image Processing, Vol. 4, No. 4, April 1995.
[2]
R. Puri and K. Ramchandran, “PRISM: A New Robust Video Coding Architecture based on Distributed Compression Principles,” Allerton Conf. Communication, Control, and Computing, Allerton, IL, October 2002.
[3]
A. Aaron and B. Girod, “Wyner-Ziv Coding of Motion Video,” Asilomar Conf. Signals, Systems, and Computers, Pacific Grove, CA, November 2002.
[4]
D. Slepian and J. K. Wolf, “Noiseless Coding of Correlated Information Sources,” IEEE Trans. Information Theory, Vol. 19, July 1973.
1Undergraduate (EECS)
2Postdoctoral Researcher

Send mail to the author : (drewc@eecs.berkeley.edu)


Edit this abstract