基於邊緣特徵之超音波與電腦斷層掃描影像對位技術

Image Registration between Ultrasonic and CT Images based on Image Edges

陳金聖(Chin-Sheng Chen)；陳俊文(Jun-Wen Chen)；黃建量(Chien-Liang Huang)

影像邊緣 ； 廣義霍夫轉換 ； 正規化相關係數法 ； 矩不變量 ； Image registration ； Image Edge Normalized Correlation ； Normalized Cross Correlation ； Generalized Hough Transform ； Moment Invariant

臺北科技大學學報

46卷1期（2014 / 06 / 01）

1 - 14

繁體中文

本文針對四種不同的影像對位方法進行分析，評估影像邊緣特徵應用在超音波與電腦斷層掃描影像對位的可行性與演算法之間的差異性，由於上述兩種醫學成像原理不同，造成無法直接進行影像對位。因此利用影像前處理，改善兩影像問邊緣特徵不一致而無法直接對位問題。再將兩種影像之邊緣特徵提取並進行對位，本文選取正規化相關係數法（Normalized Cross Correlation, NCC）、影像邊緣正規化相關法（Image Edge Normalized Correlation, !ENC）、廣義霍夫轉換（Generalized Hough Transform, GHT）及矩不變量（Moment Invariant, MI），比較其不同的對位演算法問精度與對位速度之效能分析及應用於實際案例的可行性。

In this paper, four kinds of different image registration methods to carry on analysis, evaluate image edges feature to apply the difference between the feasibility and the algorithm of Ultrasonic and Computer Tomography, because of these two different medical imaging principles, can't be directly caused by image registration. Therefore, the use of image processing, improved edges feature between two images can't be directly inconsistent alignment problems. Then two kinds of image edges feature extraction and image registration, the paper selected: (1) Normalized Cross Correlation (NCC), (2) Image Edge Normalized Correlation, (IENC), (3) Generalized Hough Transform (GHT), (4) Moment Invariant (MI), comparing their different alignment algorithm between accuracy and performance analysis of the feasibility and applied to practical cases.