臺灣博碩士論文加值系統

　　本論文提出一個基於解析式分□技巧的快速彩色影像切割法。首先我們藉著保留輸入資料的一些不變特徵而得到一個二□的解析式分群公式。然後運用此二群分群技巧，反覆地將資料一分為二，最後再用合併的技巧將相似的群合併為較大的群。如此以達到多群分□的目的。在前述分割階段裡，我們根據輸入影像量化過後之顏色，先在彩色空間上形成大略的數群。然後在合併階段時，則根據二種□的相似度量來判斷那些□該合併。當分□結果產生後，每一個彩色像點是根據其和各個□中心靠近的程度，以及和周圍空間連續性的考理來進行分類。我們也對其他不同的彩色影像切割法進行比較。實驗結果顯示，本論文所提出的方法其執行速度快，而且在各個不同的彩色座標系統皆有良好的切割結果。

　　In this thesis, a fast segmentation algorithm for color images based on analytical clustering techniques is presented. A two-class analytical clustering technique which clusters a multi-dimensional input data set into two classes by preserving some invariant features is proposed first.The technique is then applied repeatedly to handle multi-class case using split-merge concept. In the splitting phase, the quantized colors of the input image are clustered and the number of cluster is detected roughly. In the merging phase, two kinds of similarity measures are used to decide whether the merging should be taken or not. When clusters have been formed, cluster centroids are then used as prototypes. Each color pixel is then classified according to its color difference to the generated prototypes. Region connectivity is also used in classifying pixels. Attempts have also been made to compare the performance of the proposed algorithm with other existing algorithms. Experimental results indicate that the proposed algorithm with is fast and it segments images well in many color coordinate systems.