The main research theme is to develop multi-scale methods for image processing and to design efficient algorithms for image enhancement and feature detection in computer vision applications. The focus is both on the mathematical theory in the framework of non-commutative harmonic analysis and on the applications in 2D/3D image analysis.
About the theory we investigate the properties of signal representations associated with continuous frames parametrised by Lie groups, as for example the shearlet transform for 2D signals and its generalisations to higher dimensions. About applications, we are interested in algorithms for feature extraction, as edges, corners and blobs. In particular we construct local invariant descriptors, as the SIFT, able to detect directional informations of the data. A recent line of research is the application of these multi-scale methods to space-time signals, like the videos, where the time coordinate plays a different role with respect to the spatial coordinate.