標題: 三度空間腦部結構校準
Alignment of 3-D Geometric Models of Brains
作者: 蔡明倫
Ming-Lune Tsai
譚建民
Jiann-Mean Tan
資訊科學與工程研究所
關鍵字: 影像校準;ICP演算法;image alignment;ICP algorithm
公開日期: 2001
摘要: 在醫學影像的領域中,經常會比對來自不同物體模型的資料,以比較分析兩者之間的差異,因此需要一個校準方法,找出兩組物體的關連性,並且將它們對齊在一起。Besl與McKay在1992年提出了ICP(iterative closest point)演算法,可以應用在不同幾何結構的物體上,藉由尋找兩組物體間距離最近的對應關係,計算出一組最小平方的幾何轉換矩陣,反覆縮小兩組物體間的mean square distance,最後得到最佳的幾何轉換矩陣,將兩組物體對齊在一起。在本文中介紹ICP演算法做為校準方法,以人體腦部MRI腦部影像為實驗對象,從兩組不同的MRI影像中擷取出腦部白質的幾何結構,將兩組不同的腦部白質結構校準,顯示校準的結果,並且比較在不同情形下校準的差異。
In medical image processing, there exists a need to align two object models in order to compare the difference. For the two object models, we want to find the relation between them and to know how to superimpose one object with the other. Besl and McKay proposed the ICP(iterative closest point)algorithm in 1992, which was a general-purpose, representation-independent method to align different object models. ICP algorithm iteratively minimizes the distance between two object models and computes a least square transformation matrix. We introduce the ICP algorithm and use it to align the 3-D geometric models of brains. Experiment shows the result of alignment of different brain models.
URI: http://140.113.39.130/cdrfb3/record/nctu/#NT900394080
http://hdl.handle.net/11536/68607
顯示於類別:畢業論文