“…The main challenge is that for each picks we have two equations and 5 unknowns (in 2D, three map derivative terms, one surface slope misfit and one initial ray direction to source since the other one to receiver will be constrained by the offset at the surface).This can be done for example by setting a system of equation using consecutive neighbouring picks in the x,z and direction on which we will had some additional constraints to link them such as a local constant stationary map derivative. We have indeed a similar situation with respect to angle domain migration (Montel and Lambaré, 2011). The computation of the tomographic rays will require a nonlinear optimization and they do not systematically satisfy Snell-Descartes considering the dip observed in the common specular offset migrated section.…”