環場影像之接圖方法

A Method of Registering Full View Images

楊人豪(Ren-Hu Yang)

指導教授 ： 林啟芳

元智大學/工程學院/電機與資訊工程研究所/碩士(1998年)

摘要

本論文提出手持相機拍攝多張、多圈影像接合的方法。我們在一個實景中利用數位相機對於這個實景擷取多張、多圈重疊的環場實景影像,並提出的特徵點尋找方法,以影像四個頂點的座標乘上轉換矩陣後所算出的尤拉距離(Euclidean distance)。取代原先以特徵點乘上轉換矩陣後,所算出的尤拉距離作為誤差值,以確保多張影像在計算配對點時能有正確的特徵點配對,以及降低配對點的計算時間。接著我們將找到配對點的影像warp到圓球的座標系統,並計算出兩兩相鄰影像間的轉換矩陣,並利用低錯誤合併(low error merge)的方法來防止錯誤傳遞。我們以實驗結果來驗證本篇論文所提出方法的優越性。

關鍵字

影像接圖 ； 特徵點 ； 點對應 ； 全景影像

並列摘要

In this study, a novel method of registering multiple image sequences is presented for constructing a panoramic view. Those images are taken using a digital camera in a hand-held way. Due to the variation of intensities of input images caused by taking pictures at different positions and time, a normalization process is executed to adjust the intensities of each input image. Feature points are detected for each input image by adopting the interest operator [5], and three moment invariant values are calculated for each detected point, which are used as a basis for matching purpose. We also develop a selection method that can be utilized to solve the correspondence problem between two adjacent images. Each image and the associated feature points are mapped to a spherical coordinate system to improve the accuracy of the registering result. The proposed method has advantages of high stability in choosing best matching pairs and achieving accurate registering result, which are demonstrated by experimental results.

並列關鍵字

Image Registration ； Feature Selection ； Point Pattern Match ； Panoramic View

參考文獻

[1] A. Watt, and M. Watt, Advanced Animation and Rendering Techniques, Addison Wesley, New York, U.S.A., 1992.

[5] C. F. Lin, and R. J. Weng, ”A method for registering real view images,” 1998, R.O.C.

[6] E. B. Barrett, P. M.. Payton, N. N. Hang, and M. H. Brill, “General methods for determining projective invariants in imagery,” Image Understanding (CVGIP’91), January 1991, pp. 46-65.

[8] J. Flusser, and T. Suk, “Pattern recognition by affine moment invariants,” Pattern Recognition Society, Vol. 26, No. 1, 1993, pp. 167-174.

[9] J. W. Hsieh, C. C. Chiang, A. Huang, and T. Cheng, “Edge-based image registration for panoramic images,” Proc. of IPPR Conference on Computer Vision, Graphics and Image Processing, Taiwan, R.O.C., 1998, pp. 288-295.