Physical Models and Machine Learning for Scientific Imaging

Julien Mairal - Institut National de Recherche en Informatique et en Automatique - Inria, Grenoble (introduction S. Giffard-Roisin)

Deep learning has revolutionized image processing and is often considered to outperform classical approaches based on accurate modeling of the image formation process. In this presentation, we will discuss the interplay between model-based and learning-based paradigms, and show that hybrid approaches show great promises for scientific imaging, where interpretation and robustness to real-world degradation is important. We will present two applications on super-resolution and high-dynamic range imaging from photographic bursts, and exoplanet detection from direct high-contrast imaging.