- 作者: 曾定章; 呂長洋; 韋淑貞
- 作者服務機構: 中央大學資訊工程研究所
- 中文摘要: 在本論文中,我們提出了一個利用紋理對映(texture mapping)的方法來合成臉部表情動畫。首先,我們利用3D數化儀對石膏像構建一個一般化的3D臉部模型;然後,根據一張特定的2D正面臉部影像來重塑(recasting)出個人的3D臉部模型。接著,利用兩階段式(two-part)的紋理對映技術,將此2D臉部影像貼至重塑後的臉部模型上求得一靜態模型,並且建立臉部影像與個人臉部模型間的對應關係;這些對應資訊將被保留在此靜態模型中以便後續利用。在紋理對映時,我們會先把2D臉部影像做非線性的拉寬,而這個動作是隱含在紋理對映的公式中。再來,我們利用參數式的臉部動畫模式(facial animation)來改變個人臉部模型以產生許多不同表情的動態模型;並藉由靜態模型中的對應資訊,我們將2D臉部影像重新貼至動態模型上,以解決動態模型之複雜的重新對位問題。最後經由陰影(shading)與打光(lighting)等處理程序來合成不同表情的影像序列。本系統適合於模組式影像壓縮(model-basedimage compression)及交談式作業系統之圖形介面的應用。
- 英文摘要: A 3D facial animation system based on texture mapping is proposed. A generic face model is firstconstructed by digitizing a plaster bust using a 3D digitizer and then recast by using a 2D front-viewfacial image to obtain a personal face model. The facial image is mapped onto the personal face modelto generate a static model. A parameterized facial animation is applied to the personal face model toproduce sequences of dynamic models with various facial expressions. These dynamic models are simplymapped by using the facial image based on the mapping information of the static model. Finally, sequencesof facial frames are synthesized with lighting and shading. The system is complete and suitable for model-based image compression and graphic interfaces of operating systems.
- 中文關鍵字: facial animation; texture mapping; model-based image compression
- 英文關鍵字: --