Stochastic Modeling of the Spatiotemporal Wavelet Coefficients and Applications to Quality Enhancement and Error Conceal
- PDF / 1,173,830 Bytes
- 12 Pages / 600 x 792 pts Page_size
- 70 Downloads / 162 Views
Stochastic Modeling of the Spatiotemporal Wavelet Coefficients and Applications to Quality Enhancement and Error Concealment Georgia Feideropoulou Signal and Image Processing Department, ENST, 46 rue Barrault, 75634 Paris Cedex 13, France Email: [email protected]
´ Beatrice Pesquet-Popescu Signal and Image Processing Department, ENST, 46 rue Barrault, 75634 Paris Cedex 13, France Email: [email protected] Received 1 September 2003; Revised 9 January 2004 We extend a stochastic model of hierarchical dependencies between wavelet coefficients of still images to the spatiotemporal decomposition of video sequences, obtained by a motion-compensated 2D+t wavelet decomposition. We propose new estimators for the parameters of this model which provide better statistical performances. Based on this model, we deduce an optimal predictor of missing samples in the spatiotemporal wavelet domain and use it in two applications: quality enhancement and error concealment of scalable video transmitted over packet networks. Simulation results show significant quality improvement achieved by this technique with different packetization strategies for a scalable video bit stream. Keywords and phrases: wavelets, spatiotemporal decompositions, stochastic modeling, hierarchical dependencies, video quality, scalability.
1.
INTRODUCTION
Video coding schemes involving motion-compensated spatiotemporal (2D + t) wavelet decompositions [1, 2, 3] have been recently shown to provide very high coding efficiency and to enable complete spatiotemporal, SNR, and complexity scalability [4, 5, 6]. Apart from the flexibility introduced by the scalability of the bit stream, an increased robustness in error-prone environments is possible. Unequal error protection of such kind of bit streams is easily achievable, due to the inherent priority of data. These features make scalable video methods desirable for video transmission over heterogeneous networks, involving, in particular, packet losses. In most cases, however, if packets are lost, an error concealment method needs to be applied. This is usually done after the inverse transformation, that is, in the spatiotemporal domain. There exists a plethora of error concealment methods of video, most of them applying directly to the reconstructed sequences (for a comparative review, see [7]). Approaches exploiting the redundancy along the temporal axis try to conceal the corrupted blocks in the current frame by selecting suitable substitute blocks from the previous frames. This approach can be reinforced by introducing data parti-
tioning techniques [8]: data in the error prediction blocks are separated in motion vectors and DCT coefficients, which are unequally protected. This way, if the motion vector data are received without errors, the missing blocks are set to their corresponding motion-compensated blocks. However, the loss of a packet usually results in the loss of both the motion vectors and the DCT coefficients. So, many concealment methods first estimate the motion vectors associated with a missing block usi
Data Loading...
