MPEG-2 Extension Of The SHVC Base Layer

In the cable digital television industry, the support of decoding UHD(ultra high definition) set-top box decoded by HEVC(High Efficiency Video Coding) listed, forming a situation that MPEG-2, H.264/AVC and HEVC are used at the same time.In order to support set-top box which only supports MPEG-2 decoding, most of the TV channels still broadcast SD(standard definition) stream decoded by MPEG2, some TV channels also broadcast HD(high definition) stream decoded by H.264/AVC, and some individual UHD TV channels broadcast stream decoded by HEVC. Although this solution can solve the problem of compatibility between terminal devices, but most of the TV channels are not allowed to use this solution because of the limited bandwidth. The HEVC Scalability Extension(SHVC) standard using external means to provide base layer with non-HEVC bitstream, but the mpeg-2 decoding is beyond the scope of the standard. So using the SHVC solution still can’t solve the problem of compatibility.In order to adapt to the diversity of video terminal equipment, and reduce the bandwidth as much as possible, this paper extends the SHVC. We design the encoding and decoding framework of the extended SHVC with MPEG-2 as basic layer and HEVC as enhancement layer, and then two solutions about how to reduce the computational complexity of video coding are proposed. The first scheme is proposed for adaptive motion estimation search method to solve the problem of large computation of motion estimation. Another scheme through the establishment of fast mode decision model, solves the problems of high complexity of prediction unit(PU) mode decision of enhancement layer.Reducing the motion estimation search range can effectively reduce the overall motion estimation calculation. The motion vector should be mapped from base layer to enhancement layer, according to the key technology used in motion estimation between basic layer and enhancement layer. An adaptive search range can be obtained based on the motion vector mapping results. The adaptive search range reduces the number of search points and reduces the amount of computation of motion estimation.Fast mode partition model uses the properties values extracting from base layer as input and PU partition mode in enhancement layer as output, to early terminate most PU partition modes. The model can avoid calculating the cost of all partition modes. Fast mode partition model is built using the decision tree classification algorithm in machine learning, and the use of machine learning technology can improve the accuracy of prediction. Compared with the traditional method, fast mode partition model can reduce the computational complexity obviously, and the partition results have a high hit rate.This study is mainly used in video coding standard transition period, and provide a theoretical basis for accelerating the update video terminal equipment and promote the further development of digital TV.