Decoder picture buffer reduction based effective reference frame selection algorithm for multiview video coding

Abstract

To achieve compression efficiency for multi-view video coding (MVC), Joint Video Team (JVT) uses the hierarchical B picture (HBP) coding structure which based on MPEG-4 AVC/H.264 codec. The HBP structure develops a new compression tool inter-view prediction to reduce residual of the data in motion estimation. However, the exhaustive search for inter-view and temporal prediction coding by all reference frames will cause the encoder to take high computational time and large decoded picture buffer size. To solve above problems, this paper proposes an effective reference frame selection algorithm for motion estimation in MVC prediction coding. The proposed algorithm uses the relationship of between the motion degrees and its prediction direction to select the prediction reference frame in inter-view or temporal domain. The proposed architecture can save the buffer requirement in decoder, because of the decoded buffer only stores the MBs in the referenced region. The experimental results show that the encoding time for the main-view and the auxiliary-view are saved up to 33.56%, and 51.24%, respectively. The decoded picture buffer reduces to 47%. The proposed method not only achieves the low computational complexity but also reduces the decoded buffer size while maintain high quality.