Résumé:
The main purpose of this work is the fusion of multiple images to a single composite that
offers more information than the individual input images. We focus the approach within
a variational framework. First, we present the most basic variational model which is the
Poisson editing and follow it up by Osmosis. Osmosis is a transport phenomenon that
is omnipresent in nature. It differs f rom d iffusion by th e fa ct th at it al lows nonconstant
steady states. Then we study a proposed modification t o t his model t hat i s c alled jointvariational
Osmosis that makes the overall term non-convex. The minimization of this
new non-convex model gives plausible image data fusion.
We minimize it using the inertial Porixmal algorithm for non convex optimization algorithm
(iPiano), we apply the resulting minimization scheme to solve multi-modal face
fusion, color transfer and cultural heritage conservation problems.
Comparing this result with famous models visualy or quantitatively using error mesures
shows the superiority and flexibility of this method.
Keywords: Image fusion, Variational image fusion, Osmosis model, drfit-diffusion, non-convex optimization, gradient descent algorithms, proximal algorithms
