Variational Patch-Based Sparse Dictionary Learning Model for Image Reconstruction
Stanislas Ducotterd , EPFL

Seminar • 2024-08-08

Abstract
We propose a synthesis-based model for image reconstruction that takes strong inspiration from older dictionary learning methods. Unlike many other works based on dictionary learning, our model minimizes a closed-form objective and guarantees a convergent iterative image reconstruction method. We will present the different contributions that allowed our model to be competitive with other stable methods.