9.3 Shape Coding

The binary and grey scale shapes are normally referred to binary and grey scale alpha planes. Binary alpha planes are encoded with one of the binary shape coding methods (to be explained later), while the grey scale alpha planes are encoded by motion compensated DCT similar to texture coding (e.g. H.263). An alpha plane is bounded by a rectangle that includes the shape of the VOP, as described in the formation of VOP, in Section 9.1.3. The bounding rectangle of the VOP is then extended on the right-bottom side to multiples of 16 16 pixel macroblocks. The extended alpha samples are set to zero. Now the extended alpha plane can be partitioned into exact multiples of 16 16 pixel macroblocks. Hereafter, these macroblocks are referred to alpha blocks, and the encoding and decoding process for block-based shape coding is carried out per alpha block.

If the pixels in an alpha block are all transparent (all zero), the block is skipped before motion and/or texture coding. No overhead is required to indicate this mode since this transparency information can be obtained from shape coding. This skipping applies to all I-, P- and B-VOPs. Since shape coding is unique to MPEG-4 (no other standard codecs use it), then in the following sections we pay special attention to various shape coding methods.