It is clear from previous sections that the cost for computing the gradient of the objective function in the original shot-profile domain
is at least twice the cost of a shot-profile migration, because to compute the perturbed wavefields, the background
wavefields are required. Because minimizing the objective function requires a considerable number of gradient and function evaluations,
image-space wave-equation tomography in the conventional shot-profile domain seems to be infeasible for large-scale 3-D applications,
even with modern computer resources.
To reduce the cost and make this powerful method more practical, we extend the theory of image-space wave-equation tomography to the
generalized source domain, where a smaller number of synthesized shot gathers are used for computing the gradient.
We discuss two different strategies to generate
the generalized shot gathers, i.e., the data-space phase-encoding method and the image-space phase-encoding method,
both of which can achieve considerable data reduction while still keeping the necessary kinematic information for
velocity analysis.