利用小波變換與顏色字串之類似郵票檢索

Using Wavelet Transform and Color Strings for Stamp Retrieval

我們的郵票檢索方法可以很有效的檢索出類似的郵票。首先，我們將每張郵票先作二級小波變換的LL2：粗略近似的圖像取出，再將其尺寸標準化為60×60像素，再用8條規則（利用RGB三原色之相對關係）將每張郵票轉換成顏色字串，並儲存於資料庫。其次，我們利用字串比對進行比較要檢索的郵票的顏色字串與儲存於資料庫的顏色字串。最後，排名前10名相似的結果將被檢索出來。我們的系統同時保有：1.基於內容的郵票檢索系統（基於相似性檢索）和2.基於文本的郵票檢索（非常迅速和成熟）兩者的既有優勢。

In this paper, we demonstrate a system for efficient retrieval of stamps images under different conditions. First, we used each stamp to process a two levels wavelet transform, and then we used the LL2: rough approximation of the image. Next we normalized each LL2 image to 60 x 60 pixels. Second, eight rule (using the relationships between the three primary colors of the RGB color model) were used to convert each stamp into a color string, and the color strings were stored in the database. Third, we compare the color strings of stamp images with color strings stored in the database using a color string comparison method. Finally, the ten most 10 similar stamps were shown. Our system maintains the advantages of both content-based (based on similarity searching), and text-based retrieval (rapid and mature) systems.