ウェーブレット局部突出特徴を利用したコンテンツベースの画像検索

DOI Web Site Web Site 参考文献20件
  • Sum Ybngqing
    Faculty of Information and Computer Science, Keio University
  • Ozawa Shinji
    Faculty of Information and Computer Science, Keio University

書誌事項

タイトル別名
  • Content-based Image Retrieval by Combining Salient Points in a Wavelet Domain

この論文をさがす

抄録

We proposed a new image retrieval algorithm with partial indexing capability, which emphasized the local features of images by combining the salient points in a wavelet domain. These points representing sharp parts of images such as edges were detected. Then using a spatial orientation tree structure, they were combined into the Low-Low (LL) frequency subband to emphasize the local properties of the image. The wavelet coefficients in the modified LL subband and their color moments were then saved as image feature vectors, which both preserved the necessary image content and yet had enough discriminating power for image retrieval. The experimental results demonstrated that our image retrieval algorithm significantly improved retrieval accuracy, computational cost, and storage space of feature vectors in a general image database, and that partial retrieval performed much better in retrieval accuracy and robustness.

収録刊行物

参考文献 (20)*注記

もっと見る

キーワード

詳細情報 詳細情報について

問題の指摘

ページトップへ