Branched inter layer prediction structure for reduced bandwidth distribution of SHVC streams

Scalable video coding allows to efficiently distribute a number of different versions of the same video. The total bit rate is lower compared to transmitting all streams independently using a simulcast approach. However, using scalable video coding comes at a cost compared to transmitting a single high quality bit stream. In this paper, an encoding structure for the enhancement layers in Scalable High Efficiency Video Coding is proposed, which reduces the overhead compared to single layer streams. This is achieved by using the base layer as a reference for multiple enhancement layers, exploiting the inter layer prediction efficiency of dyadic scalability. This way, certain layers are not being used as a reference for the higher layers. Omitting these intermediate layers on the access network reduces the bit rate of the highest quality scalable bit stream with 22.7% on average. Furthermore, this approach reduces the decoder complexity, since fewer layers have to be decoded.