摘要
本論文中，我們提出一個以汽車外表形狀為基礎，並且建立在序列影像上的汽車辨識方法，以解決在擷取動態影像時因角度偏差而影響結果的正確性。在所提出的方法中，當輸入一段汽車的動態影像時，以影像的曲率比例空間圖(Curvature Scale Space)中所有越零點(Zero Crossing)的σ最大值，作為從序列影像中取得最接近側面影像的依據，將σ最大值較大的數張影像作為查詢輸入影像和影像資料庫資料比對，由此我們就可以以最少數量的輸入影像，來取得較佳的比對結果。
關鍵字
參考文獻
[2] A. Pentland, R. Picard and S. Sclaroff, “Photobook: Content-based manipulation of image databases”, IJCV, Vol. 18, pp. 233-254, 1996.
[3] P.M. Kelly and T.M. Cannon, “Experience with CANDID: Comparison algorithm for navigating digital image databases”, In Proceedings of the 23rd AIPR Workshop on Image and Information Systems: Applications and Opportunities, Vol. 2368, pp. 64-75, 1994.
[4] J. P. Eakins, M. E. Graham, and J. M. Boardman, “Similarity retrieval of trademark images”, IEEE Multimedia, Vol. 5, pp. 53-63, 1998.
[5] F. Mokhtarian and A. Mackworth, “Scale – based description and recognition of planar curves and two-dimensional shapes,” IEEE Trans. on PAMI, Vol.8, pp. 34-43,1986.
[6] Y. Rui, A. C. She and Thomas S. Huang, “Modified Fourier descriptor for shape representation- a practical approach,” Proc. IEEE Int. Conf. on Image Proc., pp22-23,1996.