Image retrieval based on contourlet transform and local binary patterns

A novel medical image retrieval algorithm based on texture is proposed. The contourlet transform combines non-separable and directional filters banks, and has multiscale and directional properties. The marginal distribution of contourlet transform coefficients is modeled by generalized Gaussian density. It is used for texture feature extraction in transform domain. Uniform local binary patterns have good rotation invariance. It extracts texture feature in spatial domain and his retrieval time is short. A texture feature extracting algorithm combined statistical features of the contourlet with block-based uniform local binary patterns is proposed further. The two texture feature were extracted in spatial domain and in transform domain, which are complementary. A database of medical images was retrieved by this algorithm. The result shows it can achieve a high precision of retrieval.