In this paper, we focus on the problem of compositing and
rendering a multimedia scene containing ordinary (possibly synthetic)
2D/3D objects together with 3DV objects encoded in one
of the new representation formats mentioned above. More specifically,
we propose a graphics architecture1 for multimedia platforms
supporting the seamless composition of ordinary 2D/3D
media objects together with 3DV objects containing depth information.
Since we also aim at supporting multi-view displays and
at achieving real-time performance, the proposed architecture is
based on layered-depth-image (LDI) [4] and it is implemented in
GPU via OpenCL.