基於人類視覺系統特性之視訊編碼

Abstract

本論文主旨在於利用人類視覺特性設計視訊編碼中的位元配置方式，以期達到較佳的視覺品質。論文中提出的方法著重於人類視覺中的早期視覺處理，並由於低階視域的運作行為較為一般性，所以依低階視域特性設計預測失真的公式。在視訊編碼中，通常視訊複雜度分析是設計位元配置的核心考量。在本論文中，視訊複雜度進一步分解為視覺複雜度與編碼複雜度。視覺複雜度直接影響流量控制機制，在視覺重要性較高的地方分配較多的位元，在可以承受較大失真的地方配置較少的位元。論文中並利用SSIM作為以感覺為基礎的客觀失真測量方式來評量所提出的基於視覺系統設計之流量控制機制。在H.264 JM7.6上的實驗結果顯示，論文中提出的方法與JM7.6中的參考流量控制機制比較之下，提案方法在所有測試案例中皆有較好的表現，能夠達到較佳的視覺品質並降低所使用的位元數。

This paper proposes a video bit allocation adopting perceptual model of human visual systems for better visual quality. Since the regions which attract human attention are most likely related personal experiences and are different from person to person, we do not apply the approach that decompose a video sequence into foreground and background representations. Our proposed algorithm focuses on human early vision processes and formulates a distortion measure based on low-level vision behavior because of its generality. The proposed algorithm evaluates video complexity by visual complexity and coding complexity. The visual analysis directs the rate control model to assign more bits to the regions with visual importance, and on the contrary, to assign fewer bits to the regions that could tolerate larger distortion. The proposed visual-based rate control algorithm is evaluated using a perceptual-based object distortion measurement called structural similarity index (SSIM) which approximates the perceived image distortion. Experiments based on H.264 JM7.6 shows that in comparing to the original rate control in JM7.6, The proposed method has the better performance with higher SSIM numbers and lower bitrate in all test cases.