In theory you can estimate it, but in practice this is usually planned by trial and error.

A practical method is to fly the drone straight upward at a slow, constant rate while recording video. Then pull still frames from the footage at different times, stitch them, and see which combination gives the look you want. If the ascent rate is constant, frame time gives you relative height: e.g. frames at 5s, 10s, 20s, and 60s during a 60s ascent correspond to 5/60, 10/60, 20/60, and 60/60 of the total climb.

Exact calculation is difficult because real scenes are 3D, not flat. If the scene were mostly flat, you could compare the photo to a map, identify matching points, and estimate distances and viewing geometry. With known object sizes and map distances, you can approximate height and line-of-sight relationships using basic right-triangle geometry.

So: exact latitude/longitude, height, and angle for every shot are generally not solvable precisely from the finished effect alone. The workable approach is to use a map for rough planning, then refine with test footage and frame selection.