主觪評価値の最大化に基づく動画像の時空間レート制御とその符号化への応用

摘要

ブロードバンドネットワークの普及や携帯端末等の高性rn能化により，時間や場所を問わないネットワークサービスrnが増加し，様々な動画像通信サービスが展開されている．しrnかしながら，動画像のデータ量は本質的に大きく，ネットrnワーク帯域の改善や庄縮符号化アルゴリズムの進化の一方rnで、ユーザ数やトラフィック数の増加により．%Since the bandwidth required for real-time video communication cannot always be available on a network, the number of frames per second and the average bit amount for a single frame should be controlled suitably to obtain the highest possible video quality in a restricted bandwidth. To alleviate the problem, the authors have developed a spatio-temporal rate control technique that maximizes an estimation of MOS. From a practical point of view, however, this approach offers little freedom in coding-parameter tuning and in the frame rate control. Although improved techniques have been presented for these difficulties, a detailed encoding algorithm under the spatio-temporal rate control has not been investigated. This paper thus proposes a novel spatio-temporal rate control and video coding techniques based on our MOS estimation and frame rate control techniques, and it shows their validity and effectiveness. Some applications of the proposed technique are also discussed. We have determined that 75% of videos encoded by the proposed technique give almost equal or better subjective quality compared with the conventional fixed-frarne-rate approach.