Study On Key Techniques Of Region-based Features Image Retrieval

With rapid advances in communication and multimedia computing technologies, accessing mass amounts of multimedia data is becoming a reality. The Content-based Image Retrieval (CBIR) technology can exactly meet the needs mentioned above.Several key techniques in region-based image retrieval are introduced in detail. Also an approach based on multi-dimension space and multi-feature for improving on retrieval efficiency is presented. An energy feature based on both the low frequency bands and the high frequency band of wavelet transform in the LUV color space is extracted,and according which a new energy space is builded as a segmentation space,and then the FCM clustering algorithm is used to segment the images . In the FCM clustering algorithm,the energy value is quantified to 64-level,then which is used as clustering feature vector. The initial clustering seeds and clustering block numbers is determined by the gray histogram ,which is efficient to the improvement of iteration speed. Then the visual features are extracted from all regions,and the features between the query image and the target images are matched on many to many relationship and the single feature distance between the regions is obtained according to the minimal weighted mean criterion. After normalized ,the image comparability has been obtained.