| Summary: | 碩士 === 國立東華大學 === 資訊工程學系 === 101 === In recent years, thanks to the well-developed facial image processing technology and the prevailed usage of hand-held image capture devices, more and more applications for facial processing and syntheses have attracted the users. For example, virtual plastic surgery and makeup reproducing are two typical applications. For these two applications, many software on the market generally require cumbersome manual operations and the synthesizing results are limited to only some trivial visual effects for fun. To develop an automatic solution for virtual plastic surgery and makeup reproducing, the first issue is how to track the facial features automatically and accurately. The second important issue is to enlarge the diversity of synthesizing effects. Aiming at the former issue, this thesis first proposes a facial feature tracking method by improving the robustness and accuracy of Active Shape Model (ASM) tracking. To diversify the synthesizing results, we propose a skin-color adjusting method and a similarity-based pixel blending method. Given a sample image of an interested model face with the desired facial shape and facial makeup, our proposed system can automatically emulate both the charm and makeup of the model face without distorting the identifiable personal facial appearance of the user. The simulation results show that our facial tracking method outperforms the conventional ASM tracking and the proposed system for virtual plastic surgery and makeup reproducing is easier to operate and produces more natural and fruitful synthesized visual effects comparing to other existing systems.
|
|---|
