發明
美國
13/174,501
US 8,805,038 B2
用來評估化療反應與早期偵測的紅外線影像之影像對位演算法LONGITUDINAL IMAGE REGISTRATION ALGORITHM FOR INFRARED IMAGES FOR CHEMOTHERAPY RESPONSE MONITORING AND EARLY DETECTION OF BREAST CANCERS
國立臺灣大學
2014/08/12
影像對位演算法是要將經過雙波段對位後的多時間點紅外線乳房影像形變為同一幾何結構,使相對應之組織移動到相同位置以取得更多有用的資訊。演算法大約分為三個步驟,首先利用Harris角落點偵測演算法(Harris corner detector)偵測出目標影像(target image)及欲對位之來源影像(source image)中的角落點,以及分析Harris矩陣找出血管中線之交叉點。接著利用手動選擇的方式建立兩張影像中角落點及血管交叉點的相互關係,已形成一組應用於薄板仿樣法(Thin Plane Spline, TPS)之相對應點(corresponding points)。最後利用Nelder-Mead單體法(simplex method)來修正來源影像中控制點的位置,並使得共同訊息(mutual information, MI)達到最大值,以得到最佳化對位結果。 The proposed image registration algorithm was composed of three steps. First, the corner points on the heat pattern were detected by Harris corner detector algorithm and the intersections of blood vessels were found via the Harris matrix. Second, the relationships of the corner points and the intersections of blood vessels were established between the source image and the target image manually to get a set of corresponding point for Thin Plane Spline (TPS) model. Third, the Nelder-Mead simplex method was used to modify the locations of control points on the source image via maximizing the mutual information (MI) value. In the mean time, optimal and sub-pixel registration could be achieved.
產學合作總中心
33669945
版權所有 © 國家科學及技術委員會 National Science and Technology Council All Rights Reserved.
建議使用IE 11或以上版本瀏覽器,最佳瀏覽解析度為1024x768以上|政府網站資料開放宣告
主辦單位:國家科學及技術委員會 執行單位:台灣經濟研究院 網站維護:台灣經濟研究院