2D/3D hybrid video coding based on motion compensation

This paper proposes a 2D/3D hybrid video coding technique based on motion compensation (MC). The key idea of this technique is to adaptively apply 3D(spatio-temporal) and 2D(spatial) orthogonal transformations to “macroblocks” in video signals. In this technique, the 3D DCT is selected for regions where MC works well, while an arbitrary-shaped 2D DCT coding is applied to the rest of the regions. Compared with the conventional motion-compensated interframe coding technique which basically uses only two frames in encoding a single frame, the coding efficiency can be improved with the proposed method because it takes advantage of information from more frames or a longer coding delay. A theoretical analysis together with experimental evaluations of the proposed technique are given to show the improvement of the efficiency over the conventional MC+DCT video coding technique.