透過您的圖書館登入
IP:18.226.187.24
  • 學位論文

利用 Face Appearance Model 模擬臉部外觀變化

Modeling Facial Ageing Process with Face Appearance Model

指導教授 : 李忠謀
若您是本文的作者,可授權文章由華藝線上圖書館中協助推廣。

摘要


本研究提出一個模擬因年齡增長所造成之臉部外觀變化的方法。本研究利用 face appearance model 為基礎,以同時解釋臉部外觀的形狀與灰階值變化。為了提高臉部模型的獨特性,利用 fuzzy c-means clustering 將 training set 中的所有人依臉部外觀相似度分群,每群分別建立一組 group face appearance model,以模擬每群特有的臉部外觀變化。接著,透過同一個人於不同年齡的各組參數,建立個別的 ageing function,以描述年齡與臉部外觀變化的關連性。利用 ageing function,求出各個年齡所對應之模型參數後,便可模擬因年齡增長所造成之臉部外觀的變化。實驗中,利用不同分群方式建立 group face appearance models,並比較各種 group face appearance models 模擬臉部外觀變化的結果與原始影像的誤差值。實驗結果顯示,利用臉部外觀相似度將所有人分群,每群分別建立 group face appearance models 的方法,確實能模擬各群中較為獨特的臉部外觀變化,且能提高模擬結果與原始影像的相似度。

並列摘要


This thesis proposes a method for modeling of the facial ageing process. Face appearance model (FAM) is used to model the shape and texture variations of human faces. Fuzzy c-means clustering is applied to group faces in the training set that have similar face appearances. Group face appearance models (group FAM) are built to model the unique variations of face appearance within each group. Ageing function is defined by the model parameters of each individual in different ages. The simulated results are obtained by applying the most typical model parameters of the simulated age to group FAMs. Experimental results showed that using group FAMs can better simulate facial ageing process than the usual face appearance model in terms of both visual perceptions and mathematical similarity measures.

延伸閱讀