應用於H.264/AVC之維持視訊品質與低複雜度的頻率域畫面縮小方法

Quality-preserved and low-complexity frequency-domain down-sizing method in H.264/AVC

柯佳佑(Ke, Jiayou)

指導教授 ： 陳自強

國立中正大學/工學院/電機工程學系/碩士(2012年)

摘要

隨著智慧型手機、平板電腦快速崛起，在攜帶型多媒體裝置上進行影片欣賞逐漸普及；然而攜帶型多媒體裝置上受限於體積而僅具備小尺寸螢幕，在視訊播放資料之前，需先將視訊畫面尺寸縮小成該多媒體裝置的播放尺寸。目前常見的畫面尺寸縮小方法大約分成兩種：(1)先進行反轉換並解碼成空間域畫面尺寸後，再進行空間域畫面尺寸縮小轉換的「全部解碼流程」；提供攜帶型多媒體裝置的系統保有最完整畫面資訊之優點，卻也消耗大量的計算量與儲存空間 (2)整合反整數轉換與畫面尺寸縮小轉換演算法的「簡化解碼流程」；簡化了過去反轉換與畫面尺寸縮小轉換的合併流程，雖然改善過去攜帶型多媒體裝置在全部解碼流程中存在運算複雜度高與記憶體使用量大的負擔，卻也付出了參考畫面資訊因為不再具備完整畫面資訊，導致移動補償產生誤差，造成連續畫面間的錯誤傳播效應。
在本論文中所提出的頻率域畫面尺寸縮小轉換的方法於H.264/AVC的平台上實現，與傳統簡化解碼流程的方法相同，都是將解碼流程中的整數反轉換與畫面尺寸縮小轉換流程合併，參考畫面一樣不再具備完整畫面資訊所以存在連續畫面的錯誤傳播效應；本論文更進一步提出一個利用係數補償方式，可以在簡化解碼流程中更降低運算量，採用此係數補償方式的簡化流程解碼在長寬各縮小比例(1/2)的計算量大約可以節省40%的轉換複雜度，考慮到使用簡化解碼流程會因為執行移動補償產生誤差造成連續畫面的錯誤傳播效應致使畫面品質下降；於此，可以在解碼的流程上採用部份的全部解碼流程與部份的簡化解碼流程，結合全部解碼保有最完整畫面資訊與簡化解碼流程運算複雜度低之優點，與過去使用簡化解碼流程的頻率域畫面縮小方法比較，兩者的運算複雜度接近，又可以改善固有的連帶錯誤傳播效應，延後錯誤傳播發生，以達到提昇畫面品質的目的。

關鍵字

運算複雜度 ； 反轉換 ； 次取樣 ； H.264/AVC ； 畫面尺寸縮小 ； 視訊品質 ； 視訊壓縮

並列摘要

Nowadays, many portable devices have the capability of watching TV programs. However, some of these mobile devices are restricted to low-resolution display screens. Accordingly, frame sizes of videos need be downscaled before displaying. Generally, the down-scaling methods can be classified into two categories. (1) Full-resolution approach: A video stream is decoded first and then down-scaled to meet a displaying frame size. This approach takes a lot of memory units and computational complexity. (2) Reduced-resolution approach: Inverse transforms and down-scaling are integrated together to minimize computational cost. This approach usually yields a drift error owing to frequency-domain coefficient truncation and motion compensation using small reference frames.
In this thesis, a low-complexity frequency-domain down-scaling method which effectively integrates inverse integer transforms and down-sizing is proposed in H.264/AVC. In order to decrease a drift error, an adaptive mechanism is embedded in the proposed method which adaptively employs the full-resolution and reduced-resolution approaches. As compared to the conventional approaches, the proposed method at the reduced-resolution mode can minimize the computational complexity up to 40% at a down-scaling ratio of 2:1. Accordingly, we can compromise the computational complexity and picture quality at a Group of Pictures (GOP) when some frames are performed by the full-resolution approach and the other frames are executed by the reduced-resolution approach. Restated, the proposed method can fairly preserve picture quality and lower computational complexity at a demand of video decoding and downscaling operations.

並列關鍵字

frame-size downscaling ； frequency domain ； computational complexity ； H.264/AVC ； inverse transform ； down-conversion

參考文獻

[2] W. Zhu, K. H. Yang, and F. A. Faryar, “A fast and memory efficient algorithm for down-conversion of an HDTV bitstream to an SDTV signal,” IEEE Transactions on Consumer Electronics, vol. 45, no. 1, pp. 57-61, Feb. 1999.

[3] A. Vetro and H. Sun, “Frequency domain down-conversion of HDTV using an optimal motion compensation scheme,” International Journal of Imaging, System and technology, Oct. 1998.

[4] A. Vetro and H. Sun, “Generalized motion compensation for drift reduction,” Proc. VCIP98, SPIE, vol. 3309, pp.484-495, Jan. 1998.

[14] D. Marpe, T. Wiegand, and S. Gordon, “H.264/MPEG4-AVC fidelity range extensions: tools, profiles, performance, and application areas,” IEEE International Conference on Image Processing, vol. 1, pp. 593–596, Sep. 2005.

[17] 夏夢麟，應用於視訊解碼之低複雜度反轉換與次取樣，國立中正大學博士論文，民國一百零一年二月。