以模型基礎之下臉部追蹤的方法研究-主動外型模式

Abstract

影像對應在電腦視覺裡常常被拿來應用，主要是在討論特徵點之間的對應關係。針對影響對應,近幾年來發展快速,從Active contour model, Active shape model, 到Active appearance model。此篇論文中，將探討主動外型模式方法(Active shape model)如何追蹤影像，並且將其延伸處理三維影像資訊，並討論會遇到的問題，以及提供有效的解決方法。藉由此方法，希望能夠應用在三角網格上的重建,將來在數學操作上，給予重大的幫助。

Image correspondence is a fundamental problem which discusses the relation of features between images of the same scene. Various approaches have been proposed to achieve this goal. In this thesis, we discuss the method of Active Shape Model, a straight forward approach for tracking the correspondences between images. The active shape model has been shown to be a practical and robust solution for matching features in 2D images. However, applying this technique to the matching of features in 3D point clouds still fail to produce convincible results. In this thesis, we will review the active shape model by splitting into the preprocessing and run-time stages. discuss the variation in each step, and its extension to the 3D point cloud.