An Efficient Post-processing Using DCT Domain Projections Onto Convex Sets
- Cite this paper as:
- Yim C. (2006) An Efficient Post-processing Using DCT Domain Projections Onto Convex Sets. In: Campilho A., Kamel M.S. (eds) Image Analysis and Recognition. ICIAR 2006. Lecture Notes in Computer Science, vol 4141. Springer, Berlin, Heidelberg
Post-processing methods using projections onto convex sets (POCS) have shown good performance for blocking artifact reduction. The iterative operations in POCS require infeasible amount of computations for practical real-time applications. In this paper, we propose an efficient non-iterative post-processing method using DCT domain POCS, namely DPOCS. In DPOCS, the inverse DCT and the forward DCT need not be performed by performing the low-pass filtering (LPF) in the DCT domain. Through the investigation of LPF at each iteration in the conventional POCS, the k-th order LPF is defined that is equivalent to the first order LPF with k iterations. By combining DCT domain filtering and the k-th order LPF, we define k-th order DCT domain LPF. The k-th order DCT domain LPF is performed only once to have the equivalent performance to the conventional POCS method with k iterations. Simulation results show that the proposed DPOCS without iteration gives very close PSNR and subjective quality performance compared to the conventional POCS with iterations, while it requires much less computational complexity. If we take into account typical sparseness in DCT coefficients, the DPOCS method gives tremendous complexity reduction compared to the conventional POCS method. Hence the proposed DPOCS is an attractive method for practical real-time post-processing applications.
Unable to display preview. Download preview PDF.