Content-Based Image Retrieval using Texture Spectrum Histogram and Latent Semantic Indexing
Tong Qin · Bulletin of Science and Technology · 2006
A novel texture spectrum approach that can describe variations of local pixels′ gray values is proposed.The texture spectrum histogram feature extracted from the image was used in image retrieval.The feature is low computation,moreover,experiments with it showed a good retrieval accuracy.Latent semantic indexing(LSI) is an useful technique which has been used for full-text retrieval for many years.It is used in image retrieval in this paper.A selection approach of approximation matrix's rank is also proposed.Image retrieval experiments with LSI showed that the retrieval accuracy was raised.