Ooh. Nice question. I wonder how we could prove that with a GIS?
I'd probably do a cell-by-cell comparison of the left-hand part of the image. If the lighting pattern was exactly the same for the upper/lower parts, with the exact same cell values and cell sizes, I'd conclude that the lower part was just a direct copy of the upper.
I'd use the RasterCellValueCalculator transformer in FME to do that, but there must be other GIS functionality that could do the same. ArcGIS? QGIS? Anyone?
Would be easier if we had the two separate images, instead of one fused together.