應用於高解析視訊之可適性降取樣編碼架構

摘要

本論文提出一有效的編碼架構，針對高解析視訊做壓縮編碼，主要分為兩個部份，首先參考降取樣編碼架構，研究降取樣率的最佳化來提升編碼的位元率-失真表現，再來，論文進一步考慮複雜度的問題以利系統實際實現。

降取樣編碼架構是本論文主要參考的方式之一，在傳統視訊編碼器前先做降取樣的動作，再於解碼器之後將重建畫面升取樣回原始解析度，不同於參考文獻中的降取樣編碼架構使用固定的降取樣率，本論文提出一隨著視訊內容及位元率而改變之降取樣率，能使得位元率-失真表現比固定降取樣率來得好，在高位元率有2dB以上的增益；與傳統H.264編碼方式相比在中低位元率有2dB到4dB的增益。

在複雜度方面，由於文獻資料較少針對不同畫面解析度做討論，論文先觀察在不同解析度下H.264各編碼工具的效益變動，最後提出在可適性降取樣編碼架構下針對不同解析度選擇不同的編碼參數，在視訊品質幾乎不變下能大大節省編碼時間，實驗結果顯示整個所提的編碼架構相對於傳統H.264編碼方式能減少九成以上的編碼時間，並保持在小於、近似固定降取樣率架構之複雜度。

　

Adaptive Down-sampling Coding Scheme for High-definition Video

Abstract

High-definition (HD) video that provides enhanced viewing experience is becoming increasingly popular. However, HD video requires large transmission bandwidth and computation complexity. This thesis proposes an efficient coding scheme for HD video by utilizing sub-sampling technique. First, we propose a down-sampling coding scheme with adaptive resolution-ratio to achieve better rate-distortion performance for video signals. And then, the computational complexity is greatly reduced by skipping un-necessary encoding modes.

The Down-sampling coding, which sub-samples the image and encodes the smaller sized images, is one of the solutions to raise the image quality under insufficient rates. An adaptive resolution-ratio for down-sampling coding is utilized instead of fixed resolution-ratio. The optimum resolution-ratio is derived based on the models of down-sampling distortion and coding distortion. Simulation results show that the rate-distortion performance of adaptive resolution-ratio is higher than H.264 by 2 to 4 dB at low to medium rates.

The complexity analysis of encoding tools for video at different resolutions has not been addressed much. This work analyzed quality gain of high complexity tools at different resolutions. Based on this analysis, we propose an adaptive encoding configuration scheme to reduce the computation complexity by skipping modes with low quality gains. As simulation results are shown, with almost the same rate-distortion performance, the proposed scheme further reduces complexity of the down-sampling coding. Compared with H.264, it has 90% complexity reduction at low to medium bitrates.