Wide-aperture, prestack depth migration requires application of challenging and time-consuming velocity analysis and depth focusing, collectively referred to here as depth focusing. We present an approach to depth focusing (1) using a detailed starting velocity model obtained by a 1-D transformation of the first-arrival times, followed iteratively by (2) interactive analysis of the common-image gathers, (3) computation of coherency attributes of the wavefield in the depth domain, and (4) 2-D traveltime tomography to update the background velocity model. We employ two interactive method of migration velocities refinement. In the first method (similar to the common-midpoint velocity spectrum approach), by picking the residual, velocity updates directly from the common-image gathers. In another method (analogous to the common velocity stacks), we pick the velocity updates from the areas of maximum coherency in depth sections that are migrated using rescaled traveltime maps. Both types of migration velocity picks, optionally combined with the first arrivals, are inputs for a 2-D traveltime inversion scheme that uses either the infinite-frequency or a finite-bandwidth approximation. This flexible and versatile depth focusing approach is implemented for several prestack depth migration algorithms and illustrated on an application to a real, ultrashallow seismic data set. The technique resolves overburden velocity variations and facilitates reliable high-resolution reflection imaging of a paleochannel that was the target of the study.