In this project, we aim to develop new spatiotemporal prediction tools, as an alternative to the traditional frame-by-frame intra and inter prediction techniques. Since motion estimation is one of most computationally demanding tasks on hybrid video codecs, the alternative spatiotemporal techniques can give an important contribute while developing low complexity/high performance video codecs. Particularly, it was demonstrated that not only the Least-squares prediction can achieve competitive results when compared to the traditional motion estimation, but also that there exists a strong correlation between edges in space and object trajectories in the time domain. Based on these observations, we believe that both the spatial and temporal redundancy can be exploited with similar approaches, namely using three- dimensional coding units.
|Start Date: 01-09-2014|
|End Date: 01-09-2016|
|Team: Sérgio Manuel Maciel de Faria, Nuno Miguel Morais Rodrigues, Luis Alberto da Silva Cruz, André F. R. G., João Miguel Pereira da Silva Santos|
|Groups: Multimedia Signal Processing – Lr, Multimedia Signal Processing – Co|
|Local Coordinator: Sérgio Manuel Maciel de Faria|