Object region based color image retrieval integrating multi-features

A novel indexing and retrieval methodology based on object region that integrates color, texture, and shape feature is proposed in this paper. The major problem in developing this method is the difficulty of segmenting object region from the image. To overcome this difficulty, dual wavelet transform (DWT) is used to detect salient object region in the image. Gabor filter is used to extract texture feature and three normalized inertias are used to describe shape feature. A similarity formula is developed to index image integrating three features. In order to further improve the retrieval speed, a fuzzy clustering algorithm is developed and employed to significantly save query time without compromising the retrieval precision. Experimental results demonstrate that this method has better performance in terms of retrieval efficiency than the method using single feature or total image s features, and has faster speed than other similar retrieval system.