PDA

View Full Version : Best Scanning Resolution?



rmd-photography
26-Aug-2007, 13:56
I am doing some 4x5 scans on my Epson 4490 and having to stitch. What is typically the best resolution to scan at to give sufficient detail worthwhile of a 4x5 negative?

nicol_verheem
26-Aug-2007, 21:05
I typically stick with 600 dpi until I'm sure I nailed the "exposure"; then I (mostly) rescan at 1200 with the same settings.

If I absolutely love the image and want to print very large, I scan at 2400dpi, but that really strains my PC. 4000dpi/48b gave a 2.2GB image, which I've only done once to show off, but Photoshop CS2 couldn't even save the resulting file !

It's a compromise of quality vs speed; the large files quickly get very unpractical to work with. And most test will tell you that your scanner does not show any more quality from a native 4000dpi scan, than from a 2400 dpi scan interpolated to 4000dpi. So at least save yourself the time on the scan.

Do use 48 bit if your software allows that; especially if you use multi scanning, I think the dynamic range if more of a issue than the resolution on these scanners.

--nlv

Daniel_Buck
26-Aug-2007, 21:25
I scan my 4x5s at 2400dpi with a 4990, do my levels and curves, then downsize to my print size at 300dpi. (I do as Nicol suggested, scanning in a quick lower res version first just to more accurately check the levels)

Not sure if this is noticeably better than scanning closer to the native resolution you want to print at, but downsizing from a larger resolution will give you more control over the look of your pixels and help reduce digital scan noise a little. That's the work flow I've gotten into. 4490 should be about the same as my 4990. I scan B&W 16 bit, if you are scanning color then 2400dpi might be a bit of a hassle to work with and store, due to large file sizes. So you might be better off scanning at 1200dpi if you are working with color. 1200dpi from a 4x5 will give you roughly 6000x4800 pixels, sufficient to print 8x10s and even 11x14s.

This might be a good 'archival' resolution to keep your digital files at, for the majority of your work. Special negatives maybe save them at a higher res (2400dpi or even 4800dpi, though I am still not sure that scanning any higher than 2400dpi on a flat bed produces any different result on the final print).

Ted Harris
27-Aug-2007, 11:07
In our scanning workshops we always recommend scanning at around the scanner's tested native resolution or a teeny bit over. 2400 hits that point for the 4990/4870/700/750. There is no reason to do two scans, a prescan and final at your selected resolution should be fine. Your prescan will be at ~ 150 but will give you sufficient resolution to see the detail you want. For the OP, the problem with the 4490 is that it is a step below what is normally considered "entry level" for scanning 4x5 and larger film. I can't tell you if the resolution is the same as that of the 4990 since we have never tested it. My guess is that, while the resolution may be ok, the DMax will not be as good as that of the 4990, et. al. If you only shoot/scan 4x5 occassionally then you can probably geta way with the method you are using now for smaller prints. If, you want to be able to produce top quality prints for clients/exhibitions in sizes larger than 8x10 then you should consider a different scanner, one that will scan a full 4x5 at a minimum.

paulr
27-Aug-2007, 11:17
i scan at the scanner's hardware resolution (4800ppi) at 16 bits and have the software sample it down to 2400. this averages 4-pixel blocks down to one pixel, rather than throwing out alternating pixels and scan lines. the result is a bit less noise and less likelihood of aliasing.

I sample down to the print driver's native resolution at final size after cropping, rotating, spotting, and a round of capture sharpening. i sample down to 8 bits after all the tonal adjustments are done.

John Curran
27-Aug-2007, 14:59
Boy! Scanning seems to be the most controversial aspect of this art form. The instructions that came w/ my nikon coolscan recommend scanning at a resolution based on the output size needed for the print or end use. So i would interpret that to mean 300 dpi and in the case of an 8x10 image 2400x3000 pixels. That's obviously inconsistent with what's being said here, where is my understanding flawed?

Also I've read that re-sizing an image softens sharpness and requires sharpening, however, doesn't sharpening add noise to a digital file?

Is the nikon recommendation specific to the 35mm format and not applicable to LF?

I'm confused.

john

sanking
27-Aug-2007, 15:43
Boy! Scanning seems to be the most controversial aspect of this art form. The instructions that came w/ my nikon coolscan recommend scanning at a resolution based on the output size needed for the print or end use. So i would interpret that to mean 300 dpi and in the case of an 8x10 image 2400x3000 pixels. That's obviously inconsistent with what's being said here, where is my understanding flawed?

Also I've read that re-sizing an image softens sharpness and requires sharpening, however, doesn't sharpening add noise to a digital file?

Is the nikon recommendation specific to the 35mm format and not applicable to LF?

I'm confused.

john

John,

Your understanding is not flawed, merely limited in the range of possibilities available.

Many people prefer to scan at the largest practical resolution for archiving. That way you don't have to scan again later if you want a larger print. If you scan at 360 ppi at 8X10 and then later decide that you want to make a 16X20 print your only option is to re-scan. If you scan at the largest practical resolution and archive the file you have the option of re-sizing at any point in the future. Digitizing an analog image results in some loss of sharpness so sharpening is pretty much obligatory whatever you do.

Sandy King

Ken Lee
27-Aug-2007, 16:43
One thing you might find interesting is to explore the Red, Green, and Blue channels, and find out which ones gives the sharpest results. My guess is that the Green channel is usually sharpest, because typical CCD sensors contain more Green pixels than Red or Blue. (This is relevant for scanning B&W images only).

When my scanner was tested a few years back, the Green channel gave the sharpest results.

Depending on your scanner software, you may be able to control which channel(s) make up the scanned image. For example, VueScan (http://www.hamrick.com) has an option called "Make Gray From", and the user can choose Red, Green, Blue, Auto, or InfraRed.

If you use a staining developer, you can maximize the effects of grain masking and exposure compensation, by scanning in the color which most closely resembles the stain.

John Curran
28-Aug-2007, 13:42
Thanks Sandy

John

paulr
28-Aug-2007, 14:18
BThe instructions that came w/ my nikon coolscan recommend scanning at a resolution based on the output size needed for the print or end use. So i would interpret that to mean 300 dpi and in the case of an 8x10 image 2400x3000 pixels. That's obviously inconsistent with what's being said here, where is my understanding flawed?

scanning at print resolution is just a guideline for minimally acceptable results. the trouble with (besides the archiving issue) is that anything you do that requires pixels to be interpolated (like rotating to straighten out the scan, sharpening, etc.) is going to be happening at printable resolutions. you're also more likely to see aliasing effects from film grain or fine detail forming interference patterns with the scanning grid.

when you oversample, you push all these artifacts into a frequency range that will be discarded when you downsample to your printer's native resolution.

it's similar to why it makes sense to scan and work at 16 bit word length even thought the printer drivers only interpret 8 bits. in this case, it's to get rid of the artifacts of tonal manipulations.


Also I've read that re-sizing an image softens sharpness and requires sharpening, however, doesn't sharpening add noise to a digital file?

if it isn't done carefully, it can exaggerate existing noise, and make an image look harsh and artificial. but done right it looks completely natural. sharpening methods are tied closely to the file's resolution and the size of the final print, so they're usually chosen on a per-print basis. it's a complex subject; blatner and fraser give great advice in their Real World Photoshop book.

sanking
29-Aug-2007, 09:25
One thing you might find interesting is to explore the Red, Green, and Blue channels, and find out which ones gives the sharpest results. My guess is that the Green channel is usually sharpest, because typical CCD sensors contain more Green pixels than Red or Blue. (This is relevant for scanning B&W images only).



Actually this does have some relevance for scanning color negatives if you plan to convert the image to a B&W image. If you look at a scan of a color negative in the R, G, and B channels the combination of R+G (dropping the B) will almost always give slightly better sharpness and less grain. And of course there is a lot you can do to change the tonal values of the scene with creative control of the channels.

Sandy King

Gordon Moat
29-Aug-2007, 11:56
One issue is that blue filtration passes less light than red or green. This can sometimes cause the appearance of blue channel noise. However, it really does depend upon the scanner. Many scanners have equal numbers of red, green, and blue pixels. A few also have a row of white pixels. There is also the matter of green filtration passing more light, which is sometimes why the green channel might seem cleaner or sharper. Another consideration is optics in the scanner; if this was truly an APO optic, then each channel would be equally in focus; I have seen cheaper scanners display one or two channels out of focus.

Ciao!

Gordon Moat
A G Studio (http://www.allgstudio.com)

John Curran
29-Aug-2007, 13:28
blatner and fraser give great advice in their Real World Photoshop book.

thanks paul

john

sanking
30-Aug-2007, 19:01
One issue is that blue filtration passes less light than red or green. This can sometimes cause the appearance of blue channel noise.
Ciao!

Gordon Moat
A G Studio (http://www.allgstudio.com)

Is this really so?

Some years ago I did a lot of in-camera separations with B&W film, using sharp cutting Red, Green and Blue filters. The correction for the Red and Green filters with all of the panchromatic films I used was about +3 stops for Red and Green filtration, about +2 stops for Blue filtration. I took this to mean that panchromatic film was more sensitive to Blue light than to Red and Green by about one full stop, and that Blue filtration passes *more* light than Red and Green, not less.

Sandy

Gordon Moat
30-Aug-2007, 20:35
Sensors (CCD) in scanners do not respond the same way as panochromatic film. I would think from your example that orthochromatic film would require another alteration?

Anyway, CCDs are less Blue spectrum (wavelength) sensitive than IR sensitive. To compensate, the thickness of the Bayer filtration can sometimes be optimized to even out the spectral response overall. However, not all trilinear CCDs are optimized, and some perform better than others. So it does not always appear.

To complicate this more, it is often difficult to discover what sensor, or sensor manufacturer, imaging CCD is in a scanner. Even the optics are tough to figure out for manufacturing source. So I don't know this 100%, though in some discussions with engineers, and by combined observations of many scanners, blue channel noise in lower cost scanners has been more apparent. If anything, this is partially an argument for buying better scanners.

I have a few white papers on this, including one study done for Sinar in 2001. There is a mention of the Beer-Lambert law that explains the transmittance variation due to filtration thickness, though oddly enough it affects red transmittance worst, so red filtration tends to be much thinner than green or blue. The effect of thinner red filtration is often enhanced spectral sensitivity over the other filtered pixels in the sensor. There is a bit more on dark current noise affects, which for some reason have less impact towards the IR end of the spectrum, and more effect towards ultraviolet. So again, it is not strictly blue filtration, but if there will be problems there are a few reasons they might show up in a blue channel more than red or green.

I hope that all makes some sort of sense. I have been suffering from exhaustion and heat stroke for several days, so I am probably not as sharp as usual.

Ciao!

Gordon Moat
A G Studio (http://www.allgstudio.com)