Well I can't draw a diagram. But let me try to explain ddepth migration with words. What we do is we use velocity analysis to come up with a velocity model. Now, by using the ddivergence theorem we can come up with Green's functions that allow us to express the seismic field recorded on a measuring surface as the field at some point within the Earth volume. I realice that is a confusing sentence, but all it means is that we have an equation which allow us to take our seismic data recorded at the surface (i.e. At z=0) and transform it to the data that we would have recorded if we put our geophones at some depth in the Earth (z=z1 or something). Now we must assume that our data is a recording of the upgoing field only to do this. Then we can use an estimate of the source field(i.e the wavelet) and use basically the same equation to calculate what the source field is at the same depth. So we almost have our recipe. We take our data at z=0 and back propagate it to depth z1, we forward propagate our source field to depth z1 as well. Now if there is a reflector just below z1, then the zero lage of the cross corellation of our newly propagated wave fields will be non zero. If there is not a reflector just below z1 then the zero lag of the cross correlation will be zero. Finally, all we have to do is step our data down through the Earth, using the velocity model and Greens functions, to successive depths z1, z2, z3 etc. At each step we take cross correlation of the two fields and we output zero lag value to the migrated output. Thats it. Sorry if its not too clear. Tough without equations.