Research On Relevance Technology Of Region-based Image Retrieval System

The amount of digital images is increasing in a surprising speed, which is of great benefit to people. At the same time people have to face how to be analyze, store and retrieve these huge amount of data effectively. This paper is about research on relevance technology of region-based image retrieval System, which includes the studies of color space quantization, region-based image segmentation, similarity computation, relevance feedback and etc.The color quantization is an important method in image retrieval, at present, color space quantization schemes are short of considering human perception. So this paper proposes a new color quantization method based on 11 kinds of color in symmetrical L*a*b* space, which can reduce the color categories and the influence of the same color with different illumination on image segmentation.As for region-based image segmentation, we segment image into regions with the model of mixture Gaussian and the EM algorithm, and propose a new method which is based on pixel feature through the average scatter degree and average divided degree and used to choose the best cluster value of K. Experiment results indicate that our method can segment images efficiently.As for image similarity computation, we present an approach based on integrated efficient region, which can reduce the adverse effect of inaccurate segmentation, and increase the weight of efficient regions and decrease the negative effect of inefficient regions in the retrieval process. Our method improves the retrieval precision, and has better performance which is validated by some experiments.As for relevance feedback, we introduce a SVM-based image query method, which make the system can guess user's interest by his choices and automatically adjusts the measurement rule to improve the accuracy. Based on above mentioned researches, we design and implement an integrated efficient region based image retrieval system, called IERBIR. In our proposed system, the feature database is constituted with SQL Server to save image features, and some function modules are included, such as feature extraction, image segmentation, image retrieval, relevance feedback and etc, so such has practical values.