如何從遙測影像中擷取有用的資訊來完成地貌辨識是本論文的研究主題。因此我們的研究內容共分為三大部分：(i) 影像強化，我們利用小波轉換的多重解析度特性，分別針對不同解析度的高頻係數，以小波收縮 (wavelet shrinkage) 去除雜訊，同時以Teager能量運算 (Teager energy operator) 強化較大區塊的邊線對比。(ii) 邊線擷取，主要在於邊線追蹤 (edge tracking) 與小波轉換的結合，利用多重解析度的高頻資訊作邊線追蹤，有效解決雜訊及邊線不連續的問題。 (iii) 以線段為特徵的影像比對，將線段轉換到不同的向量空間，以Hausdorff distance作為比對的方法，解決影像的旋轉、大小變化及位移等問題，達到可靠且有效的比對結果。 In this study, approaches of image enhancement, edge extraction, and line-based image matching for remote sensing images are proposed. The image enhancement includes noise reduction and contrast enhancement. We apply wavelet shrinkage techniques to suppress noise while preserving the sharpness of large-scale edges based on a Teager energy operator. The edge extraction contains wavelet-based edge detection and tracking. Wavelet transform provides multiresolution representation of images for robust tracking. The proposed edge detector consists of three modules: (i) starting point extraction and purgation for tracking, (ii) multiresolution gradient image generation, and (iii) multiresolution edge tracking. The image recognition approach matches line-based features using invariant Hausdorff distance. This approach matches two images and solves the problems of rotation, scaling, and translation transformations between these two images by applying the process of minimizing Hausdorff distance twice on the two sets of feature vectors.
