Transform for inter-layer prediction residues in scalable video coding

Inter-layer prediction is an important tool in scalable video coding schemes. With inter-layer prediction, the encoder can eliminate redundancy between the layers and improve compression efficiency. In this paper, the characteristics of inter-layer prediction residues are investigated and an improved transform scheme is presented. Compared to traditional schemes using DCT-II as the spatial transform, the proposed scheme introduces two additional transforms. These additional transforms are designed in a way such that they may better capture the dominant components in interlayer prediction residues while their implementation can reuse the transform modules in existing codecs. Results obtained with SHVC reference software SMuC0.1.1 show that the proposed method can improve the coding efficiency with negligible increase in the encoding complexity.