For that shot, I would never go to Photoshop unless I needed the original background of the sign. Let's say that was the case. Here's the project what I would do.
- Pick a frame where the side of the apple box was as close as it gets to square to the camera.
- Export a frame as a PSD
- Draw a mask around the edge of the apple box
- Duplicate the layer using Ctrl/Cmnd + j to get a copy with just the cut out apple box
- Change to the selection tool (v) and transform the apple box to square it up to the frame using the Ctrl/Cmnd key to move the corners
- Add the new sign elements in another layer
- Save the PSD, import it to AE and drop it in the timeline above the footage
- Select the footage layer in AE and choose Track In Mocha from the Animation Menu (or from the Effect Menu in the latest version)
- Draw a spline around the outside edge of the Apple Box and track
- Adjust the surface to include the same area that I used in Photoshop
- Check the track to make sure it was good by adding an image like the 8 X 8 grid to the surface
- Export the corner pin data, or copy or just apply directly from the Effect Control Panel to the PSD
All done, all through. Just for fun, I'll give it a try a little later today. It should take about 10 minutes.
Edit: Took me 7:44...
Here is the workflow.
If all I needed was to add a text layer with a blend mode I could have eliminated the photoshop file entirely.