在低頻寬網路環境中利用內容感知之視訊調整

Abstract

隨著無線網路的發展與可攜式隨身裝置能力的提升，視訊傳遞的應用越來越廣泛。在有限的資源和與生俱來的限制下，如何適當地調整視訊成為無線多媒體應用相關領域一個最重要且富挑戰性的議題。在這篇論文中，我們提出了一個以內容為依據的視訊調整方法，讓資源能夠有效地利用，並提高影片的視覺品質。藉由分析影片中物件的亮度、位置、移動向量與能量等屬性，配合客戶端裝置能力與關聯性統計模型取得場景中較吸引注意的部分。利用區域加權速度失真模型、位元配置與視訊調整方法，動態地從物件、畫面與GOP三個層面來做視訊串流的調整。從實驗結果可以看出所提出的方法相當實用並可達到較好的品質。

With the development of wireless and the improvement of mobile device capability, video streaming is more and more widespread applied in such environment. Under the limited resource and inherent constraints, appropriate video adaptation has become one of the most important and challenging issues in wireless multimedia application related areas. We propose a novel approach to adapt video based on content information in order to effectively utilize resource and improve visual perceptual quality in this thesis. According to the analyzed characteristics of brightness, location, motion vector, and energy features, combined with capability of client device and correlational statistic model, the attractive or interesting regions of video scene are derived. Therefore, the Region Weighted Rate-Distortion is used for adjusting the bit allocation. Video adaptation scheme dynamically adapt video bitstream through object, frame, and GOP levels. Experimental results show that the proposed scheme is efficient and achieves better visual quality.