Abstract: Aiming at the 72-dimensional HSV color feature in content-based image retrieval(CBIR), this paper proposes a new dimension reduction idea. It preserves the intrinsic merits of the original image color feature. Based on dimension reduction, it puts forward a new indexing structure to improve the performance of content-based retrieval of large image databases. Experiments show that this method is effective.

Key words: Dimension reduction, Intrinsic dimensionality, Content-based image retrieval(CBIR)

摘要： 针对图像的72维HSV颜色特征，提出了一种新的降维方法。该方法在降维的过程中充分保留了图像颜色的本征特性。在降维的基础上，建立了一个新的索引机制，并以此加速大规模图像库的基于内容检索的进程。实验证明，该方法是行之有效的。

关键词: 降维, 本征维, 基于内容的图像检索

CLC Number: 

• TP391.41