Lecturer: Barbara Pascal
The purpose of this course is to provide a cross-disciplinary view on multiscale analysis tools from mathematical modeling to concrete applications in signal processing in the broad sense, with a focus on image processing. As preliminaries, multiscale analysis will be introduced formally in a general setting, and then presented through the prism of the continuous and discrete wavelet transforms. Then, a thorough presentation of the concept of scale invariance in signal and image processing will prepare for the introduction of the texture segmentation task. Two recent frameworks for isotropic and anisotropic fractal texture segmentation will be detailed, from the fractal models to the efficient algorithmic schemes leveraging proximity operators. The two methods will be illustrated on real-world data, targeting computer-aided mammography analysis.