Deployment of wavelet transform to VOP codec scheme

Abstract

The work explains the way we deploy the wavelet transform in the object-based video codec scheme proposed in our previous research (see Tran, S.M. et al., J. Commun. and IT Sciences Association, vol.LVI, no.4, 2001). Video objects are first extracted from a conventional frame-based video sequence. By representing the video objects of interest with the so-called constrained Delaunay mesh structures, we could transmit their shapes and patterns in an efficient way: mesh-nodes are tracked forward to reconstruct their new position and therefore their shape at the latter time point; the pattern of the mesh is predicted from the dense field of backward motion vectors, then they are compensated with residual errors for a "perfect" reconstruction. In the current work, the wavelet transform is applied to encode residual errors, the most bit-consuming information. Thanks to the multiresolution character of the wavelet transform, which leads to superior energy compaction and perceptual quality of the decompressed residual errors, the method surpasses the compression offered by the block-DCT in the initial phase. The overall performance of the full scheme is encouraging.