CODING EFFICIENCY AND COMPLEXITY ANALYSISOF MVC PREDICTION STRUCTURES

摘要

Based on the idea to exploit the statistical dependenciesfrom both temporal and inter-view reference pictures formotion compensated prediction, this paper presents a systematicevaluation of multi-view video coding with optimizedprediction structures. The compression method isbased on the multiple reference picture technique in theH.264/AVC video coding standard. The advantages of hierarchicalB pictures for temporal prediction are combinedwith inter-view prediction for different temporal hierarchylevels, starting from simulcast coding with no inter-viewprediction up to full level inter-view prediction. When usinginter-view prediction at key picture temporal level averagegains of 1.4 dB PSNR are reported, while additionally usinginter-view prediction at non-key picture temporal levels averagegains of 1.6 dB PSNR are reported. For some casesgains of more than 3 dB, corresponding to bit rate savings ofup to 50%, are obtained.