Abstracting data from within alignment>stitching process

[cvhmanchester]

I am studying C19th print techniques, by scanning large impressions on paper at high resolution, 1200 ppi or photographing them. I have been using MS ICE very successfully to stitch these large (1.7 GB, 16 bit TIFF) images. It is also used for stitching mosaics of focus stacked macrophotographs. Surprisingly for an excellent free program, it is not mentioned in the stitching tutorial.

ICE takes the overlapping scans, aligns them, then stitches them. If the alignment is easy, e.g. a linear sequence of images, the stitching seems pixel perfect. Indeed the overlap area seems to have increased resolution by averaging. If alignment is not easy, e.g. a checkerboard, the averaging produces a fuzzy area in the overlap.

I wish to identify differences between nearly identical impressions in order to reveal the progressive changes added by the artist. Instead of creating an average matrix from the 2 aligned image matrices, is it possible to extract the difference matrix +/- after the alignment stage?

Since ICE seems no longer supported by MS, it is unlikely that I shall get a direct answer, unless someone has reverse engineered the program's flow.
Does anyone know of a program that will enable me to extract the data files of the aligned images before they are stitched? It would be even better if one could adjust the amount of alignment to allow for the differences between the images.
Thank you

[Manfred M]

The "old" Photomerge functionality in current versions of Photoshop does exactly that. It stacks the images that are to be merged as layers, aligns them and creates masks to show which parts of the layers are aligned and overlapped. The newer software found in software like Lightroom, Adobe Camera Raw, Affinity Photo, etc just provide the finished merged product without the "trail" that one sees in ICE or the Photomerge functionality in Photoshop do.