Active Shape Model with Applications in Facial Landmark Localization
Sun, Zhe (2016)
Sun, Zhe
2016
Master's Degree Programme in Information Technology
Tieto- ja sähkötekniikan tiedekunta - Faculty of Computing and Electrical Engineering
This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Hyväksymispäivämäärä
2016-06-08
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:tty-201604203840
https://urn.fi/URN:NBN:fi:tty-201604203840
Tiivistelmä
Active Shape Model (AAM) and Active Appearance Model (AAM) are the commonly used methods in facial landmark localization, in the past 20 years, there are a few extended methods, which are based on the classical ones, being published.
In this master thesis, the classical Active Shape Model (ASM) and Active Appearance Model (AAM) are well studied and summarized, especially ASM, which is also introduced in formulation. Two newly extended versions based on Active Shape Model are introduced and implemented through the thesis work. Stacked Active Shape Model (Stasm), which is much closer to the classical ASM, achieves a very good result on frontal face image landmark detection, so that it is the emphasis of this thesis. Besides we use Component based ASM as the comparison method, which is another Active Shape Model method based on component analysis. We performed these two methods for facial images from different situations: frontal and non-frontal images, single and group images. From the observation and data results, we show that Stasm still has room for improvement on facial feature localization. We explore the theoretical differences of these two extended versions and propose ideas for improvement in the later chapters.
In this master thesis, the classical Active Shape Model (ASM) and Active Appearance Model (AAM) are well studied and summarized, especially ASM, which is also introduced in formulation. Two newly extended versions based on Active Shape Model are introduced and implemented through the thesis work. Stacked Active Shape Model (Stasm), which is much closer to the classical ASM, achieves a very good result on frontal face image landmark detection, so that it is the emphasis of this thesis. Besides we use Component based ASM as the comparison method, which is another Active Shape Model method based on component analysis. We performed these two methods for facial images from different situations: frontal and non-frontal images, single and group images. From the observation and data results, we show that Stasm still has room for improvement on facial feature localization. We explore the theoretical differences of these two extended versions and propose ideas for improvement in the later chapters.