Are G-Buffers the same as the multi-pass render passes or is that something else?

What I am trying to do is simple. I have a picture of a wall with windows, and I have a 3D glowing sign. What I am trying to do is to have the wall be a bit illuminated by the sign and for the windows to catch the reflection of it.

The wall is not the simple straight refular wall, it has a bit of scylpting and architectural embelishment. That's why I'd like to keep as much of the original image as possible, because I don't think I can adequately and accurately model it all. Plus, the image is of the place that I don't have access to (might not even exist anymore), so all I have to work with are just a few static images of it from a particular angle.

I'd like to make this into an animated video as the end result, with the sign elements flickering on and off, and for the light on the wall/windows to accurate reflect such interaction with the flickering lights.

Does this help to clarify the situation a bit?