I have aerial photo prints on photographic paper. I want to scan these with high quality scanner.

Realistically, what is the highest resolution that I can scan at and be able to have good resolution but not waste disk space (i.e. large image file size)? There must be a point where you cannot gain much additional quality by increasing the scan resolution.

Is it possible to gain a better quality scan from negatives or transparency positives rather than prints? I would like to reduce the photo cost by purchasing transparencies and scan at a higher resolution, rather than purchasing a large print and scanning it.

From: "Petros" <petros@rock.rx>
Date: Thursday, February 21, 2002 12:33 PM

Are they colour or black and white? For colour there is not much point in going over about 300dpi and for black and white 600dpi is certainly possible.

Yes, scanning the negatives will always result in better images. However a scanner capable of scanning large format negatives will not be cheap.

From: "Jonathan Buzzard" <jonathan@buzzard.org.uk>
Date: Thursday, February 21, 2002 3:58 PM

In part this depends upon the photo - whether it is a contact print or an enlargement, the paper, etc., but a good quality normal AP print will have a resolution of well over 1000 dpi, whether it is colour or monochrome.

Yes ... but!  The film grain size can be measured in a few microns - If the print was contact - ie 'full plate' - then you will get better resolution from the negative, certainly, but whether you have a scanner up to it is a different question!

You need especially to beware of desktop colour scanners - their colour resolution is almost always worse than their grey or line art resolutions. There will be other problems too, since the scan may well not be even, adding another source of distortion to that of the camera lens.  Scanning film, rather than paper, on these is also problematic, due to the illumination method.

There are specialised scanners available for scanning film, as opposed to prints, but these do tend to be significantly more expensive than their paper equivalents.

From: "Peter Halls" <pjh1@york.ac.uk>
Date: Friday, February 22, 2002 2:30 AM

It depends entirely on what you want to use the scans for. We routinely get aerial photography scanned for use in soft-copy photogrammetric systems, and for that they are scanned at 15 or 25 microns (1700 dpi and 1000 dpi respectively). This requires scanning in a specialized scanner optimized for aerial photography. We have tested standard desk-top scanners, but the geometric fidelity is not really up to the job - and it costs a lot to get a scanner that will take a standard 9" air photo with an optical resolution in the right ball-park (resampled resolution won't do). The jump in price from a standard A4 scanner to one that will take an image 9" square is rather startling, and you're looking for a 1000 dpi optical resolution as well. We send our scans out to a bureau that has a scanner designed for aerial photographs ( they cost big money), and think it cheap at the price of about £20 per scan. A further point with desk-top scanners is that they can't handle roll material, and air photo negatives (the best material for scanning) are kept on the roll. For photogrammetric purposes, you scan either the negatives (deprecated for fear of damaging them, but gives the best results) or contact prints.

We are normally scanning monochrome photos (Antarctica tends to be pretty black and white!), but there isn't really much loss of resolution in a colour air photo - we are using specialized film, optimized for the job. I don't think it would be worth scanning either type at a higher resolution than 10 microns - although the grain is smaller than that, it takes many silver grains to produce the half-tone effects.

From: "Paul Cooper" <paul.cooper@bas.ac.uk>
Date: Friday, February 22, 2002 4:40 AM

Paul is correct, it depends on what you want to use the photos for.

I just wanted to add a couple of things to what Paul wrote.

The highest resolution you would want to use for black and white is 5 microns, the silver halide grain size is a little more than half that size. For color, you would probably want to go with 7.5 microns. However, I really can't tell a difference between a 5 micron B&W, a 7.5 micron color or just a 15 micron scan or either.

So, I tend to always go with 15 microns regardless of the film. Plus, you really have to look closely to tell a difference between 15 microns and 30 microns. Use 15 microns for precise mapping, 30 microns if you want it for GIS type work.

You would also be well advised to compress the images. I prefer JPEG with a Q15 for color, Q25 for black and white. A 15 micron JPEG Q25  b&w image will end up at around 58Mb. Uncompressed is 4x larger, around 256Mb.

Color at 15 microns JPEG Q15 is around 80Mb, uncompressed is around 800Mb.

You could also compress using something like MrSid, but personally, I think MrSid tends to clip contrast a little too much. Also, it looks to me like Sid compression introduces artifacts into the imagery.

Finally, for mapping work, stick with black and white. Much nicer contrast and a wider range of shades visible to the human eye. For GIS, use the pretty color pictures so you can ooooh and awwww.

 (certified photogrammetrist)

From: "bugzbunny" <bugz@bunny.net>
Date: Friday, February 22, 2002 10:42 AM

Thanks for your additional comments - I didn't have some figures to hand. I would take issue with you over image compression, though. JPEG is NOT suitable for analytical work, and neither is MrSID or ECW. All are fine for images to be used as backdrops, but analytically they are useless. This is because all three are lossy compression techniques that lose some of the detail that you have spent money on acquiring. By all means use LZW compression in TIFF (50% - 70% compression, depending on the nature of the image), but lossy compression means that you might just as well scan at a lower resolution in the first place. JPEG in particular is bad, because it can not only reduce detail, but it can also introduce artifacts (look at a JPEG image near a strongly contrasting sharp boundary - you will see light/dark bands parallel to the boundary). Compression at the 10% level is in general impossible without loss of detail, which is unacceptable in many photogrammetric applications. I should, perhaps, note that we are mapping terrain with no artificial features, so we are heavily dependent on small detail in our images.

From: "Paul Cooper" <paul.cooper@bas.ac.uk>
Date: Friday, February 22, 2002 11:09 AM

You have already lost some information by starting with paper prints.

It depends on what minimum resolving power you need. A 12 micron pixel (approx 2000 dpi) will have different ground pixel sizes depending on the scale of the photograph. As a rule of thumb, doubling the resolution (in dpi) gets you about the square root of two times the resolving power. For the work I am doing, with vegetation consisting of clumps ranging from 1-metre across up, I need ground pixel resolution of 0.5 metre, so that I have at least four pixels per object. Eight would be better, but the file sizes increase exponentially.

Creating diapositives loses information. Unless your scanner is very good, the best solution is to get the negs scanned on  a photgrammetric-quality scanner. If that's too expensive, diapositives are preferable to paper prints.

From: "Mike McBain" <mjm@towncrier.cc.monash.edu.au>
Date: Monday, February 25, 2002 6:31 AM

Top of Page  

Back to SURE 340 web page.

This page is maintained by Bob Burtch and was last updated on 11 April 2002.
Please send any comments or suggestions here.