隨著科技快速的發展及數位典藏計畫的實行,使得以影像內容為檢索特徵(content-based image retrieval)來搜尋相關貝類資訊已非難事。本論文提出一個以貝類影像形狀、紋理、顏色及貝類殼孔特徵為基礎的貝類檢索系統,在形狀方面採取二值化面積比對的方式,在紋理方面採取小波紋理分析,在顏色方面採取HMMD顏色直方圖分析,在殼孔方面採取索貝爾(Sobel)邊緣偵測分析,針對正面及背面的各項影像特徵擷取出其特性。透過網際網路分散式架構,使用者先上傳需要檢索的貝類圖片,經過影像的前處理,將影像去除雜訊並對貝類物件影像大小、位置及主軸方向作正規化動作後,再運用貝類影像正面特徵(殼孔)及貝類背面特徵(外形、顏色及紋理)的擷取及比對,最後依外形做為檢索比對的門檻值並利用投票法來輸出比對後的結果。由實驗結果可以得知,所提出之貝類檢索系統,具有相當不錯的檢索結果。
As the rapid development of information technology and the implement of national digital archives program, it seems to be not very difficult to search seashell information using content-based image retrieval techniques. In this study, we propose a seashell indexing system based on the extraction of image features. Three commonly used image features- shape, texture and color of seashell images and the seashell hole image feature are employed in our system. These features are implemented using the area of binary seashell object, the texture analysis of wavelet transform, the HMMD color histogram extracted from front view images, and the Sobel edges matching of seashell holes extracted from back view images, respectively. Based on the web distributed architecture, firstly a user can submit a seashell image to our web site. The system then performs the image preprocessing, normalizing and the feature extraction of the submitted image. Finally, a vote algorithm is used to classify the image into one category in our database. Experimental results show the proposed method has a good performance for seashell indexing.