區域導向並採用色鍵式形狀編碼之視訊編碼

Abstract

視訊資料通常需要很大量的儲存空間或傳輸頻寬.近年來,各種電訊服務以 驚人的速度成長中. 因為服務量的持續增加,而頻道的總容量有限,故我們 需要低位元率之視訓編碼技術.近來,區域導向視訊編碼受到很大的關注. 一個視訊區域的資料可以分為兩部份:形狀資訊和紋理資訊.形狀資訊的編 碼通常十分昂貴,尤其在低位元率的情況下.高效率的形狀鞭碼法通常很複 雜.我們認為色鍵式方法可能是一個簡單又有效率的解答.在本論文中,我 們提出一個區域導向視訊編碼法.其中採用色鍵式形狀編碼,使我們可以將 一個區域的形狀資訊與紋理資訊以同一種方法一起編碼.在分割區域時,我 們採用的主要技術是區域成長與區域合併.並且我們採用平移式區域式運 動補償.我們作了電腦模擬實驗.結果尚可但為瑧理想.使表現不理想的一 個主要原因是區域式運動補償的效率不佳.改善區域式運動補償的方法是 一可考慮的未來研究題目. Video data usually require a very large amount of storage space or transmission bandwidth. In recent years, various telecommunication services have been growing at an incredible speed. Because ofthe increasing number of services and the fact that the totalchannel capacity available is limited, low bitrate video coding techniques are needed.Recently, region-based vvideo coding has gained significantattention. The information of a video region can be devidedinto two parts, i.e., the shape information and the textureinformation. The coding of the shape information is usuallyexpensive, especially at low bitrates, The shape coding methodsthat have good efficiency are usaully complex. We find thatthe chroma-keying approach may be a simple and efficientsolution.In this thesis, a region-based video coding with chroma-key-basedshape coding is presented. Adopting the chroma-keying approach,we can code the shape information of a selected region together with its texture information using only one coding method. Thekey techniques we use for region segmentation are regiongrowing and region merging, and we adopt region-based motionestimation with translational motion. Computer simulationis conducted. The performance is acceptable but not ideal.A main reason for the nonideal performance is the inefficiency of the region-based motion estimation. It is of interest to improve the region-based motion estimation in future work.