Approximating DPI as part of image characterization

[kba]

If dpi/dpix/dpiy ist not in the EXIF metadata or cannot be trusted to be accurate, original page extent must be an input to the characterization.

[kba]

Very unlikely to be feasible since, if the pixel density isn't in the images, their physical extent won't be in the metadata or not reliable or not per-page etc.

[kba]

However bad the input, OcrdExif must be consistent. Output unified resolution (sqrt(xResolution * yResolution) and consistently set to 1 if not provided/inferrable.

[bertsky]

Output unified resolution (sqrt(xResolution * yResolution) and consistently set to 1 if not provided/inferrable.

I am in the middle of fixing #256 – do you want me to fix that, too?

[kba]

I didn't realize you were working on it, I'll leave it to you then. Have a look at kba@76181e7 for what I did until now.

[bertsky]

Sorry, neither did I! I tried to incorporate everything I saw in your commit in PR #257 ...

[bertsky]

All of the revised goals have been captured in #257 already, therefore closing.