一种基于LSI的图像语义检索技术

探讨如何将LSI技术应用于图像检索中，以实现基于语义的图像检索的技术途径，给出了一种新的图像索引方法，它使用灰色模型GM（1，1）对图像的像素值进行模型化处理，并且使用模型参数的概率分布来描述图像，在此基础上，详细讨论了将隐含主义索引应用于图像检索中的具体方法，并给出了相应的算法，最后，给出了实验结果和尚需进一步研究的问题。

An approach to semantic-based image retrieval using LSI

Latent Semantic Indexing (LSI) is a new and promising technique for information retrieval, which has been successfully used for document process and retrieval domains. Within the limits of our knowledge of publish work, only a few papers are concerned with the LSI technique for content based image retrieval. In this paper, the approach to semantic image retrieval using the LSI technique is investigated. A novel image indexing method is introduced, which uses the grey model GM(1,1) and Hilbert space filling curve to model the pixels in an image, and uses the distribution of the GM(1,1) parameters as the image representation. The method to LSI based image retrieval is discussed in detail, and the corresponding algorithms are given. The experimental results are reported and the future work is presented.