Has anybody looked at the specs. for the new Canon scanner due out in May? It will scan 35mm to 4x5 negs and slides at 2400x4800 dpi with an expected street price of $499. It also has new scratch and dust suppression technology.

I'm shopping for a scanner and don't want to drop the $1 to 3K needed for a good scanner, at least at this point. I am still dedicated to silver B&W work and just got the capability to do 8x12 inch color with a new Canon BJC-8200 (Watch Out Epson). If I really get hooked on digital, then I will spent the big $ needed to do 16x20 or larger on my computer. I want to crawl before I mortgage the house.

-- Gene Crumpler (photocrumpler@att.net), March 12, 2001

Answers

What's let these 'high resolution' flatbeds down in the past has been poor optical quality.
I looked at the Epson 1200 perfection, which claimed 'true 1200dpi optical resolution', and tested it with a resolution plate. It delivered 1200 pixels per inch alright, but the actual optical resolution was closer to 600 dpi, in other words, about 12 line pairs per millimetre.
I spent hours trying to refocus it, and came to the conclusion that the lens was just rubbish.

There's a widely-held belief that the glass platen reduces the resolution of flatbeds, but I proved to myself that this wasn't the case with the Epson, because even with the glass removed the sharpness was terrible. I can't see any good reason for not getting a sharp image through a piece of plate glass.

So, I'm skeptical that the new Canon will deliver on what its specification promises, but I'm also ready to part with my cash if it does.

-- Pete Andrews (p.l.andrews@bham.ac.uk), March 13, 2001.

Pete,

Could you expand on the method you used for testing resolution? I would like to do some of my own testing.

I'm also interested in getting a step wedge to test Dmax. Someone on the rec.photo.largeformat group mentioned some test he had run using one.

-- Marc Bergman (mbergma2@ix.netcom.com), March 13, 2001.

Hi Marc.
The resolution plate I used was a 2" square lippmann plate copy of a custom semiconductor, resolution-test, 'chrome' master. If that means anything to you I'll be very surprised.
Basically, it's a near-enough grainless plate with horizontal and vertical resolution bar patterns on it. The bars range from a spacing of 2 microns to 100 microns; equivalent to 250 line pairs per millimetre to 5 lppm.
It was only when the spacing reached 40 microns that the Epson scanner could resolve the lines (~635 dpi).

I only used such a plate because I had easy access to it, and could make a number of high-resolution contact copies.
The 'American airforce resolution test target' (type 51) would do equally well, and is easier to get hold of, but don't ask me who sells it.

-- Pete Andrews (p.l.andrews@bham.ac.uk), March 14, 2001.

Pete,

Thanks for your help.

The USAF 1951 target will work just fine. I can get one from Edmund Optics.

-- Marc Bergman (mbergma2@ix.netcom.com), March 19, 2001.

The point whether a scanner resolves or not a pair of line is not like pure black or pure white. The scanner will gradually resolve less and less of the details of the target. Therefore, there is still a little modulation left at 1200 dpi (which is 600 lines/mm). Pete did not mention the modulation limit (expressed usually in percents) he has defined as "resolving".

And this is where the computer has a decisive advantage over the silver film: it is possible to amplify this modulation. Starting from the little modulation left by the scanner, you can go back not only to the level of the original film, or even further to the original details of the scene photographed. This means that you can recover the details at much more than what Pete mentioned by carefully unsharp masking the image.

The unsharp mask is a high-pass spatial filter. It acts by amplifying the high frequency modulations of the image (i.e. the details...). By applying several times the filter with decreasing radius and increasing strength, you can rectify the modulation in the high frequencies to the level on the film.

The only difficulty here is the noise... The noise of the scanner shows up as very high frequency details (pixels), and you amplify them when you unsharp mask. Therefore, there is a limit to what you can recover. But it is undoubtedly possible to recover details up to more than 2000 dpi from a flatbed scanner with 2400 dpi optical, and this nearly whatever the lens (provided it is not absolutely horrible of course, but the Epson one is quite good in my opinion...). I will certainly buy this Canon 2400 scanner !

I can personnally get perfectly sharp images from my Epson 1200 Photo scanner with a careful mutiple unsharp masking, definetely much sharper than anything you can do by resampling 600 dpi with the same scanner. By sharp, I mean with details up to the pixel to pixel modulation, i.e. 1200dpi. The CCD noise is quite strong on this scanner, though, but you can relatively easily use Photoshop and/or multiple sampling with Vuescan to remove these unwanted signals out...

-- Pierre Kervella (ke_pi@hotmail.com), June 06, 2001.