Image retrieval with the nonlinear dimension reduction

In this paper we present a new content based image retrieval model. In order to improve the retrieval efficiency and accuracy, variance based color image quantization is presented and color auto-correlogram is constructed based on the quantized image; locally linear embedding is used to reduce the dimensionality of the feature space, it proves to play a critical part in the success of CBIR system. Relevance feedback is designed to bridge the semantic gap between the simplicity of available visual features and the richness of the user semantics. To illustrate the potential of such an approach a prototype image retrieval system has been developed and preliminary experimental results on a database containing about 1000 images demonstrate the effectiveness of the proposed model.