1: \begin{abstract} The convergence of many algorithms involving a gradient descent relies on its Lipschitz constant. When it its calculus is difficult, the backtracking rules are helpfull. When handling epigraphical constraints, proximal algorithms are the most suitable to find a solution. A choice of a differentiable or non-differentiable regularization will then lead to the choice of the optimization algorithm. In this work we compare the performances of the forward-backward and primal-dual Condat-V\~u algorithms with backtracking, for the epigraphical constrainted high dynamical reconstruction in high contrast polarimetric imaging.      

2: %\end{abstract}

3: