標題: 利用特徵點檢測法則和空間座標判定法則作部份影像辨識Image Registration Using Feature Nodes Detection and Spatial Coordinates Discretion 作者: 周永隆Chou, Yung-Long林昇甫Sheng-Fuu Lin電控工程研究所 關鍵字: 特徵點徵測;小波轉換;Feature detection;Wavelet transform 公開日期: 1997 摘要: 在這篇論文中我們利用特徵點檢測法則和空間座標判定法則作部份影像 辨識。一開始我們利用小波轉換(wavelet transform)將影像中灰階變化 大的地方找出，這些地方是我們抽取特徵點時重要的依據。當影像有雜訊 干擾時，為了防止雜訊點誤認作 特徵點，我們提出一套有效的方法來解 決這個問題。然後我們定出在空間座標上 特徵點之間的距離和角度關係 。最後我們由特徵點本身的資料和空間座標的判定法則，成功地將原始影 像的待辨認部份找出。 在我們的系統中是允許任意圖案的影像來進行辨 認，此外我們的部份影像和原始影像 之間也有了比例縮放、旋轉角度和 雜訊的干擾。 In this thesis, we use feature nodes detection and spatial coordinates discretion to build image registration. At the beginning, we utilize wavelet transform to indicate the places with great change of gray-level on age. And those places are the base to extract features. When there is noisy disturbance on the image, in order not to choose a noise-disturbed pixel to be the feature node, we propose a useful method to solve this problem. Then we define distance and angle between two feature nodes in spatial coordinates. Finally, we use the data of feature nodes and the spatial coordinates discretion to successfully find out the part waiting for our recognition on raw image. That our recognition system allows various kinds of graphs is a progress in image registration. Besides, our partial image has been scaled, rotated or disturbed by noise, compared with raw image. URI: http://140.113.39.130/cdrfb3/record/nctu/#NT860591048http://hdl.handle.net/11536/63228 Appears in Collections: Thesis