Conclusions and Future Directions
This thesis investigates novel motion anchoring schemes for highly scalable video compression schemes, with the aim of improving the temporal scalability and accessibility features. Existing video compression schemes employ block-based motion, which are unable to accurately represent motion information in the vicinity of moving object boundaries. Instead, in this thesis, we focus on “physical” motion, which accurately describes the trajectory of each pixel location. In a compression scenario, such dense motion fields are generally discarded because of their high coding cost. We employ a recently proposed highly scalable representation of motion discontinuities using breakpoints. Breaks can be used to adapt the wavelet bases in the vicinity of (motion) discontinuities; this significantly reduces the coding cost of motion fields while retaining sharp moving object boundaries in quantized motion fields.