Multiresolution Filter Based Image Database Retrieval

Multi-resolution texture features that are based on the local power spectrum obtained by a bank of Gabor filters are applied to a set of CT brain images and its retrieval performance is experimentally observed and compared. The features differ in the type of non-linear postprocessing which is applied to the local power spectrum. The following features are considered: Gabor energy, complex moments, and grating cell operator features. The capability of the corresponding operators to produce distinct feature vector for different textures is compared using the classification result. The grating cell operator gives good retrieval results. The texture detection capabilities of the operators and their robustness to rotation are also observed. The grating cell operator is the only one that selectively responds only to texture. The retrieval performance is also compared with the statistical texture features and the gabor based features outperforms.