Active Shape Model And Its Application In The Interpretation Of The Human Face

The goal of computer vision is not only to recover image structure but also to know what it represents. Especially, the image interpretation of flexible objects is more difficult due to the variation of shape. Recently, the most popular approach to resolve * such problem is building model with parameters based on statistical methods. The primary content of this thesis is about the theory of building ASM models and the approach of interpreting face using the models.Firstly, the concept of the statistical models in computer vision and the advantages of interpreting images of flexible objects using them are discussed.Secondly, I research the process of building ASM models with parameters based on statistical methods. Discuss the procedure of extracting the information in the images using Point Distribution Model and building models using Principle Component Analysis.Then, the process of image interpretation using ASM model isdiscussed. The Multi-resolution approach is used which makes the searching quicker and the system more robust.Finally, I design the system of modeling ASM model and face searching which landmark the images in the training sets, build models and search face using multi-resolution methods. The system proved good in landmarking face images, modeling ASM model automatically and analyzing and interpreting images.