The Research Of Intra Mode Coding For The Depth Maps

In recent years, 3D videos which usually consist of multi-view videos have been popular with the public. Thus, there is an urgent need for more efficient video compression technology dealing with the huge amount of video data. 3D-HEVC can do the 3D video compression, and compared with traditional compression algorithms merely on each single view, it reduces the redundancy among views and improves the coding efficiency with inter-view prediction technology by the full usage of correlation among views.The traditional HEVC intra-prediction method can't completely suitable for the intra-prediction of depth map for the current 3D video format. Therefore, 3D-HEVC brings in Depth Modelling Modes(DMM), which is more in line with the characteristics of depth map.However, the introduction of DMM greatly increases the complexity of the original algorithm.In order to reduce the time of coding algorithm, the current study makes the following improvements on DMM from the perspectives of depth map characteristics and DMM characteristics by utilizing the correlation among prediction models:Firstly, the DMM1 optimization algorithm based on intra-mode correlation is set forth in the thesis. DMM1 in 3D-HEVC needs to traverse all wedgelet modes and measure the prediction accuracy of each possible wedgelet mode in each iteration process with View Synthesis Optimization(VSO), which produce great computation cost. By studying the characteristics of depth map and DMM, the present study improves the existing DMM1 optimization algorithm, reduces the DMM search ranges and lessens the video coding time by utilizing the correlation between edge texture and prediction mode.Secondly, the DMM skipping algorithm based on inter-view correlation is put forward in present study. In the intra-prediction process of depth map, as an expansion mode of traditional intra-prediction method, DMM was added to the candidate list of intra-prediction mode. Then it conducts complete rate distortion evaluation and select optimal intra-prediction mode for the current coding block. However, when the current coding block is not applicable to DMM, the DMM search will lead to unnecessary computation resource consumption. By studying the correlation among the prediction models and among the view points, the currentthesis tries to skip unnecessary DMM prediction selectively and further reduce the complexity of video coding.To sum up, the study proposes two kinds of optimization algorithm for depth map intra-mode. It has certain reference significance for real-time encoding application.