臺灣博碩士論文加值系統

在本文中, 我們提出一個利用紋理對映方法來合成臉部表情動畫的系統.
首先, 我們利用兩張一般性的臉部結構圖來構建一個一般性的臉部模型.
然後, 根據一個特定的二維臉部影像來重塑出個人的臉部模型 ; 並且透
過紋理對映的步驟, 將此二維臉部影像貼至重塑的臉部模型上 , 並且建
立起臉部影像及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. 接著我們利用參數式的臉部動畫模式來改變靜態模型以產生
許多不同表情的動態模型; 並藉由靜態模型中的對應資訊 , 我們將二維
臉部影像重新貼至動態模型上. 最後經由陰影(shading)與打光(
lighting) 等處理來合成不同表情的影像序列.

A 3D facial animation system based on a 2D texture mapping is
proposed. At first, a generic face model is constructed using
two generic facial meshes. Next, the face model is recast by a
2D front-view facial image to get a personal face model, and
then the facial image is mapped onto the personal face model to
generate a static model. A parameterized facial animation is
applied to the static model to generate several sequences of
dynamic models with various facial expressions. These dynamic
models are also mapped from the facial image based on the
mapping information of the static model. At last, several
sequences of frames of facial expressions are synthesized with
lighting and shading.