一种适合Web图像检索的图像降维算法研究

本文对现有Web图像检索技术现状进行了归纳，分析阐明了图像降维算法在基于内容的Web图像检索技术中的地位和作用。在介绍了几种经典图像降维方法后，重点介绍了国外近来提出的基于拉普拉斯特征值映射（LE）的图像降维算法。针对Web环境下图像检索必须进行剧烈降维的特点，本文进而对基于KL变换的主成分分析（PCA）算法和基于LE的图像降维算法进行了实验分析和比较。实验结果表明：对于需要进行维数剧烈缩减的Web图像检索来说，基于LE的图像降维算法可以获得最佳的效果。

Research on an Arithmetic of Image Dimensionality Reduction for Web Environment

In this paper, the authors survey recent research of Web image retrieval, analyze and present the important effect of image dimensionality reduction on content-based Web image retrieval. After introducing several classical methods of image dimensionality reduction, the paper gives a detailed introduction on the arithmetic of image dimensionality reduction that based on Laplacian Eigenmap. Due to the characteristic that image retrieval needs drastically dimensionality reduction in Web environment, we performed several experiments to evaluate the effectiveness between the arithmetic that based on LE and the PCA arithmetic that based on KL transform. Experimental results show that the arith- metic hat based on LE give excellent performance for image retrieval in Web environment.