基於特徵點追蹤之虛擬試戴應用

Abstract

隨着網路的發達，網路購物變得越來越流行。國外的眼鏡業者爲了促進銷售，把眼鏡放到網路上販賣。然而消費者無法得知這個眼鏡是否和自己合適，這對網路行銷是一大致命傷。爲了讓消費者得知自己戴上這幅眼鏡的樣子，本研究將利用虛擬實境技術將3D的眼鏡模型合成到消費者提供的照片上。傳統的虛擬試戴技術是透過3D臉部模型和人工點選臉部特徵等方法得知人臉和鏡架的關係。不同於傳統方法，本論文研究利用特徵試鏡架和合成用之3D鏡架模型的對應關係完成虛擬試戴。如此可以避免長時間的3D建模和人工選取特徵點的困擾。

With the advance of Internet, online shopping becomes more and more popular. Foreign vendors put glasses on the Internet for sale. But a customer cannot be certain whether a particular pair of glasses will fit s/he well. In order to deal with this problem, this research uses AR technique to synthesize 3D glasses models onto the facial image provided by the customer. While traditional AR techniques obtain the fitting relation between a human face and different glasses frames by constructing 3D facial model and manually selecting the facial features, this paper accomplishes virtual try on by establishing the corresponding relation between the test and the model glasses frames via physical contact. Thus we can avoid the trouble of creating 3D model and the manual selection of facial features.