In object-based video coding, the model for the motion compensated prediction cannot be applied for the special objects and uncovered backgrounds. To reconstruct these regions, coding of the signal information in both regions is required. In this thesis, efficient coding methods for the signal information of these regions are proposed.
The special objects, which are arbitrarily shaped, can be efficiently coded by positioning the coding blocks. Block-positioning algorithm is proposed. The uncovered backgrounds can be efficiently coded by the predictions from background memory and neighboring pels. Algorithm to obtain the background memory in object-based video coding is proposed.
Simulation results show that the proposed methods code the information of these regions with less data than that in the conventional scheme.