Mathematical methods are an alternative to tackle visual perception. The central

idea behind these methods is to reformulate the visual perception components

as optimization problems where the minima of a specifically designed objective

function "solve" the task under consideration. The definition of such functions is

often an ill-posed problem since the number of variables to be recovered is much

larger than the number of constraints. Furthermore, often the optimization process

itself is ill-posed due the non-convexity of the designed function inducing the

presence of local minima