Image Retrieval Based on Combined Color with Texture Feature

CHANG Zhe,HOU Yu-qing,LI Ming-li,CHENG Tao,LIU Li-ning
DOI: https://doi.org/10.3969/j.issn.1000-1220.2012.05.042
2012-01-01
Abstract:A new method for image retrieval combining texture with shape features using wavelet is proposed,which has the property of rotation,shift and scale invariance.The rotation-normalized image is achieved by rotating the principal axis to the weighted orientation of angular moments.The adopted wavelet,which has the property of shift and scale invariance,is used to decompose the orientation-normalized image.The wavelet energy of every sub frequency band is extracted to be the texture feature of the image.Using the sub-graph rebuild the graph and then extract the shape features by Hu invariant moments.Gaussian model is used to normalize the texture and shape features.Finally,the feature combined texture with shape is used to retrieve.The experimental results show that the proposed algorithm has better robustness and precision.
What problem does this paper attempt to address?