[Medical image retrieval based on nonlinear texture features].

To extract texture features from medical images, the approaches for multi-scale complexity and multi-scale fractal dimension were proposed in this paper. The extracted texture features were used in the medical image retrieval experiments. The complexity measurements used include permutation entropy and 2D-C0 complexity. The former is used to analyze one dimensional signal, and the latter is used to analyze two dimensional one. Comparison experiments were carried out between our approaches and others' approaches in some literatures. Preliminary experimental results show that our approaches can effectively describe the texture information of medical images. Furthermore, the image retrieval results are encouraging.