Hybrid semantic and codebook mouth animation scheme for model-based coded video

Previous model-based codecs have used two distinct approaches for mouth animation, namely texture codebooks or semantic animation parameters. Each approach when used on its own suffers drawbacks in terms of a bit rate/quality trade-off. An investigation is presented into the performance of a new hybrid scheme for mouth animation in which codebooks and semantics are combined and switching between them is achieved depending on the amount of motion exhibited by the lips. Experiments show that both subjective quality and PSNR measures of the reconstructed images are improved