視訊圖框率及取樣頻率轉換之研究

Abstract

在本文中, 我們針對視訊信號格式轉換中的兩個領域, 提出數種低複雜 度, 且能提升畫質的轉換方法. 對於圖框頻率轉換部分, 我們提出像元移 動像量預選及影像分割兩種方法, 前者可減少運算量並提升畫質 , 後者 則可將運算量減至極低. 在掃瞄頻率轉換部分, 我們提出在頻譜上的內插 理論, 可免除設計複雜濾波器的麻煩, 並可達到一理想濾波器的效果. 最 後我們針對 MPEG2解碼器的輸出介面提出一個超大型積體電路的設計, 其 為解碼器的一部份. 功能是控制輸出畫面和週邊元件的同布, 畫面在螢幕 上的位置, 色訊號的內插及轉換和畫面長寬比的轉換. In the thesis, we propose several efficient frame rate conversion algorithms which can improve image quality with low computational complexity. For frame rate conversion, we propose two algorithms, one is based on MCI algorithm with pixel-level motion vector preselection, the other is besed on block MCI. The first algorithm acieves a better image quality at a reduced computation load, while the second algorithm costs even less computation and still maintains a comparable visual quality. For sampling rate conversion, we propose a transform-domain based conversion algorithm, which does not have to design a strict and complicated filter. Compared with the time-domain interpolaion filtering approach, not only the algorithm gives better performance, but it may also reduce computation load in some cases. Finally, we implement a video output interface unit for MPEG2 decoder. Its function is to synchronize the output picture to the peripheral devices and position the picture on the screen, to interpolate chrominance samples, to convert the color space from YCrCb to RGB and the aspect ratio from 4:3 to 16:9.