| Summary: | 碩士 === 南台科技大學 === 資訊工程系 === 99 === The performance and recognition accuracy improvements for image retrieval have become important research issues in recent years. In many applications of the image database retrieval, the amount of the images is too much and too similar to have a good performance. The Zernike moment technology is one of the most famous image recognition methods. Especially, Zernike moment has a high anti-rotation property for the image retrieval. However, in order to improve the retrieval accuracy in an image database, the Zernike moment method must increase its moment degree, which increases the computation time drastically. Therefore, this thesis proposes a cutting strategy based on Zernike moments, texture feature and Radon transform for an image. This thesis employs these features to reduce the retrieval time and to improve the retrieval accuracy in the image database. The experimental results show that the proposed method performs better than Liu et al. and Kim et al. methods.
|
|---|
