My contribution: If you are scanning in any document that is double sided, the bright light of the scanner will often cause the opposite side to bleed through slightly. I’ve found that one can eliminate this almost entirely if you cover the opposite side with something that is black - several sheets of black paper or black construction paper, or the plain black back of certain hardcover books.
My question: Are there any decent auto-feed photo scanners out there? Or is it better to scan negatives even though they are smaller? In which case, are there any decent auto-feed negative scanners?
My contributions:
Black behind a double sided source should help because less light travels through the back surface and back to the scanner if the reflector behind it is black and reflects little light.
Another thing that should help is the following software, which I haven’t written but can picture: you scan both faces of the source, and the software flips and aligns one of the images with the other, and then creates two different linear combinations of the two. If the amount of image bleed through is correctly compensated, the effect goes away. That is, it shouldn’t be that hard to remove the artifact of the image on the far side if you also have access to what the image on the far side looks like. Added refinements could include nonlinearity and even accommodating different penetrating effects for the different ink pigments used in the source.
Another contribution: a scanner makes a pretty decent camera for taking pictures of some kinds of objects. Obviously, leaves and printed circuit boards and studies of how uniform the thread count in cloth is all work well, because the subject matter is inherently flat and visible. Less obviously, sometimes deep objects also scan fairly well. One scanner I used would actually give a poor quality picture of the ceiling tiles if the lid was left open (how they designed the optics remains to this day a bit of a mystery to me).
Yet another: you can write software or use image analysis packages to get dimensional data and other similar things out of scanner images. For example, you could dump a box of lockwashers on the scanner and spread them out, and then analyze how uniform their diameters and concentricities are. These things depend on scanner quality and other issues, too, so you have to have statistical controls embedded in the image, including things like accurate size scales oriented in at least two directions.
Finally, the pinnacle of uselessness: remember the old Logitec ScanMan of about 20 years ago? If you hold it up with the detector vertical and the bottom pointed away from you, start scanning, and slowly rotate the little roller while you pivot around on your swivel chair, you get a panoramic shot of the room. If you turn the roller the wrong way, the panorama will be mirrored.