一個有效抵抗旋轉的影像檢索方法

Abstract

現今有許多方法用來擷取影像的特徵，影像雜湊是其中一個特徵擷取方法，最近有許多基於影像雜湊的影像檢索方法。有些方法能處理大部分的變形，然仍無法處理旋轉的問題。因此，本論文提出一個能抵抗旋轉的影像檢索方法。首先，先判別輸入影像及資料庫影像的縮放關係，再配合對應的切割區域，產生影像雜湊，最後根據影像雜湊進行影像檢索。

Nowadays there are many ways for image feature extraction; image hashing is one of these ways. Recently there are some image hashing based methods proposed for image retrieval. Some methods can treat most of image distortions; however, the rotation problem is still not solved. Thus, developing a method for handling rotation problem is necessary. In the thesis, an image retrieval method based on image hash is proposed. It can deal with the rotation problem. First, the scale rate between an input image and a database image is evaluated, and then based the scale rate; the image hash is produced according to the corresponding circular areas. Finally, image retrieval is conducted based on the image hash.