時間階層結合空間可調式視訊編碼之快速模式決策

摘要

可調式視訊編碼(Scalable Video Coding, SVC) 除了具有傳統H.264/AVC非常高之編碼效率優點外，更提升了極大的編碼彈性，主要的原因為SVC擁有時間可調性(Temporal Scalability)、空間可調性(Spatial Scalability)及訊雜比可調性 (SNR Scalability)三大特性，且均由一個基礎層(Base Layer)及數個增強層(Enhancement Layer)所構成，其中Base Layer的編碼方式類似於H.264/AVC，而Enhancement Layer除了可自行做預測與編碼外，亦利用Base Layer之編碼資訊進行預測與編碼，由於預測的來源增加，因此亦提高其運算之複雜度，對於無線裝置之低功率與即時性通信需求來說，將更有挑戰性。

本研究論文針對SVC之複雜度作分析，提出結合時間可調性與空間可調性之快速模式決策演算法(Fast Mode Decision Algorithm)以降低複雜度，本演算法係採用在同一空間階層中，不同時間階層與空間基礎層相關性不同的特性並配合空間可調性之模式特性進行預測，經實驗結果發現本演算法比採用全域搜尋(Full Search)法最多可節省達78.28%之編碼時間且PSNR只稍降0.14dB，亦即在可接受之視訊品質需求下，本研究論文所提之演算法可大幅降低系統之運算複雜度。

　

Fast Spatial Layer Mode Decision based on Temporal Levels in H.264/AVC Scalable Extension

Abstract

Multimedia applications on various devices and heterogeneous networks become popular. However, it faces the problem that different requirements, such as bandwidth constraints, CPU processing capabilities, and display resolutions, need to be satisfied simultaneously. Scalable coding is one of the solutions that can provide temporal, spatial, quality, and rate scalabilities.

Scalable Video Coding (SVC) is getting popular but exhibits a problem of high computational complexity compared with H.264/AVC single layer coding. A fast mode decision algorithm that reduces the candidate modes in motion estimation can certainly reduce its computation time. We propose an adaptive temporal level mode decision algorithm in spatial scalable video coding. In spatial scalability, we utilize the spatial candidate modes table constructed based on the statistics of real data to reduce the candidate modes. In temporal scalability, we apply the base layer mode information directly at low temporal levels. The simulation results show that the proposed algorithm can reduce computation time up to 78.28% with less than 0.14 dB video quality degradation compared with JSVM 9.12.