An Effective Image Retrieval Method Based on
Yan Dong Tang, Hong Zhang · 2014
To improve the accuracy of image retrieval meth- ods, an effective image retrieval method based on multi- features is proposed. The color feature of the image is extracted by constructing a (16:4:4) quantization scheme of image color. Perform the marked-watershed to achieve the segmentation of the image, and then extract Jan Flusser invariant moments. Gaussian model is applied to normalize the different subcharacters distance. The final multi-features similarity consists of the color similarity and the shape similarity. Experiments demonstrate the efficiency of the proposed method.