MAXIMALLY SMOOTH IMAGE RECOVERY IN TRANSFORM CODING

This paper considers the reconstruction of images from partial coeffic ients in block transform coders and its application for packet loss re covery in image transmission over ATM networks. The proposed algorithm makes use of the smoothness property of common image signals and prod uces a maximally smooth image among all those with the same coefficien ts and boundary conditions. It recovers each damaged block by minimizi ng the intersample variation within the block and across the block bou ndary. The optimal solution is achievable through two linear transform ations, where the transform matrices depend on the loss pattern and ca n be calculated in advance. The reconstruction of contiguously damaged blocks is accomplished iteratively, using the previous solution as th e boundary conditions in each new step. This technique is applicable t o any unitary block-transform and is very effective for recovering the DC and low-frequency coefficients. When applied to still image coders using the discrete cosine transform (DCT), high quality images have b een reconstructed in the absence of many DC and low-frequency coeffici ents over spatially adjacent blocks. When the damaged blocks are made isolated by the use of block interleaving, satisfactory results have b een obtained even when all the coefficients are missing.