基於時空域畫面複雜度的品質限制跳畫面演算法

Abstract

近年來最佳化視訊編碼的限制問題已日漸成為研究的目標。本論文著重在以位元率和品質限制為條件的時空域品質衡量問題。基於編碼畫面的時間域相依性，我們提出一個可調式跳畫面機制來達到好的時空域品質衡量。在給定的區間內藉由所提出的R-Q/D-Q模型去預測所有的位元率與畫面失真的組合並且挑選那些畫面使得它在時間域上內插能達到較佳視覺效果。實驗結果顯示，我們的方法雖然簡單但非常有效地能夠解決所針對的問題，並且對於快速移動的視訊品質亦能給於適度的改善。總結而言，我們的方法能夠藉由所提出的畫面選擇方式來有效地達到已編碼畫面的品質要求。

This constrained problem of optimized video encoding has increasingly been the object of study in recent years. This thesis focuses on the problem of trading temporal quality for spatial quality subject to the rate- and quality- constraints. An adaptive frame skipping scheme is proposed to achieve a good tradeoff between spatial and temporal quality based on temporal dependency between the coded frames. The frame selection scheme chooses the frames that lead to the better visual quality in interpolation by estimating all possible combinations of the rate and distortion of the coded frames within the window using the proposed R-Q/D-Q models, respectively. Experimental results show that our scheme is simple but very effective, and have moderate improvements for high motion sequences in the various subjective-based objective measurements. To conclude, our scheme can effectively achieve the quality requirement of the coded frames in spatial domain by the proposed frame selection approach.