This work is an attempt to fill the gap between high level image analysis and low level feature detectors for volumetric imaging data. It proposes both a representation model and an extraction technique for 3D spectral features. This approach is not limited to a multiresolution decomposition, but it goes a step further, rearranging frequency bands to group correlated information together. The whole process is based on minimal assumptions, inspired in theories about the human visual system. The method is completely data-driven; no prior information is used. The performance of the method is tested in a variety of volumetric imaging applications: 3D medical image segmentation (MRI and SPECT), and analysis of Ground Penetrating Radar (GPR) data and Tagged Magnetic Resonance (TMR) image sequences. The methods presented in this book can be of interest for researchers in the field of image processing in general and, in particular, for scientist and engineers working in analysis of volumetric data of any kind.