基於離散餘弦轉換之HEVC快速畫面內預測演算法研究; DCT-based Fast Intra Prediction Algorithm for HEVC

NCUIR > College of Electrical Engineering & Computer Science > Graduate Institute of Communication Engineering > Electronic Thesis & Dissertation > Item 987654321/61531

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/61531

Title:	基於離散餘弦轉換之HEVC快速畫面內預測演算法研究;DCT-based Fast Intra Prediction Algorithm for HEVC
Authors:	王琬珺;Wang,Wan-Chun
Contributors:	通訊工程學系
Keywords:	HEVC;畫面內預測;編碼單元;約略模式決策;離散餘弦轉換;HEVC;Intra Prediction;Coding Unit;Rough Mode Decision;DCT
Date:	2013-08-08
Issue Date:	2013-10-08 15:19:59 (UTC+8)
Publisher:	國立中央大學
Abstract:	近年來，許多多媒體影像視訊諸如即時視訊會議、智慧型手機及網路視頻等等，皆致力於高解析度及高品質的發展，然而伴隨著高解析度而來的是龐大的資料量，使用HEVC(High Efficiency Video Coding)可以達到比H.264/AVC更好的壓縮率，其採用了四分樹(Quad-Tree)的編碼架構，在畫面內預測中，使用了35個預測模式來達到更精準的預測，然而為了增進編碼效能，需要計算編碼單元、預測單元(Prediction Unit , PU)及轉換單元(Transform Unit , TU)所有的組合，而編碼單元遞迴重覆地切割其深度更是造成極高的運算複雜度，因此本篇論文利用影像經由離散餘弦轉換(Discrete Cosine Transform , DCT)後在頻域上的特性提出了快速演算法：於編碼單元中，計算低頻交流值(Low- frequency AC components)來決策最佳大小，提前終止其深度的切割；於畫面內預測中，藉由DCT邊界特徵來判斷預測單元的邊界方向，篩選約略模式決策(Rough Mode Decision ,RMD)及RDO(Rate-Distortion Optimization)的候選模式個數。由實驗結果可知，我們所提出的演算法平均可節省41%的時間，並在視訊畫面上維持一定的品質。 In recent years, lots of multimedia video applications aim for higher resolution and quality, such as real-time video conference, smart phones, network video and etc. The higher resolution brings more bitrate. Using HEVC can achieve significant improvements in compression efficiency compared to H.264/AVC. In HEVC, there is a quadtree-based coding structure. To improve the coding efficiency of intra frame, the prediction mode is increased up to 35. However, it has to consider all the combination of coding unit (CU), prediction unit, and transform unit for the improving coding efficiency. The CU which recursively split into 4 CUs cause quite high computational complexity. Hence, in this thesis, we proposed a fast algorithm which employed the characters in frequency domain of image transformed by discrete cosine transform (DCT): In CU, calculate the Low- frequency AC components to determine the best CU size and early terminate the split of depth in intra prediction, judge the edge orientation of PU and choose the candidate modes of rough mode decision and rate-distortion optimization by DCT edge features. The experimental results show that our proposed algorithm achieves 41% time saving on average without significant degradation of coding performance.
Appears in Collections:	[Graduate Institute of Communication Engineering] Electronic Thesis & Dissertation

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	740	View/Open

社群 sharing

Loading...