Image merging within python - based tool (maybe OT)

[Terry Reedy]

"R. Charles Henry" <trapforcannedmeatproduct at hotmail.com> wrote in
message news:8ed56b42.0211092045.1e833602 at posting.google.com...
> I'd like to ask the group's advice.
>
> My situation: I have two separate document images, both of the same
> format .png, .bmp, .jpg, or .tif. They are otherwise identical in
size
> and file characteristics, having been produced from a larger-than-A4
> document using an A4 flatbed scanner. Each image contains a region
of
> 'overlap' with the other.
>
> I need to find an open-source 'mosaicking' or 'blending' codebase,
> which will be capable of accepting both input images, and merging
them
> to reproduce an image of the original larger-than-A4 document.
>
> This code is to be used within a tool I am creating using python, &
> I'd like to avoid the neccessity of having to manually set reference
> points on the input images. The code itself that I'm searching for
> doesn't need to be python, just accessible from it.
>
> Since the input images are 'flat', I don't require such features as
> projection adjustment etc, I just need some form of basic automatic
> 'merging'.
>
> Would anyone have any suggestions? Any pointers at all would be very
> gratefully received.

Semi-random thoughts.  First, convert to bitmap.  Then minimize misfit
as function of x,y offset of centers or whatever.  Details depend on
nature of images.  With black/white documents, misfit is % mismatch.
You should standardize scanning process so proper offset is
approximately known and search for best fit within circumscribed
region.

Entirely different approach is to apply OCR to images and then match
text.

TJR