4990 Scanner Histograms

[wesjr]

Just got my 4990 last week. Every preview scan comes back with a severely underexposed histogram. I can't understand why. I adjust it like PS levels but do not like the results. Seeing more noise than I like with my slides. My slides are not underexposed.

My settings are 24 bit color and 2400 dpi resolution. Don't know what to do. Using the eye dropper tool for shadows, midtone and highlights doesn't work for me....
_________________________
Charlie Wesley
St Augustine Beach, FL

[bez]

Presumably you’re using Epson Scan, in Professional Mode? Although the histogram is useful to check clipping, I usually rely more on the appearance of the preview image. For a single frame I aim for a low contrast & slightly dark result, then fix it in PS, unless you have hundreds of slides to bash through and want an acceptable ‘Auto’ scan.
24bit colour is actually only 8bit, so you may want 48bit for your raw scans.
Can you post an example, and perhaps all the settings you're using?

[The Crofter]

I have just checked a few of my 4990 scans and the histograms are more than acceptable from both slides and prints. I do not bother with "auto anything" and adjust if necessary in CS4 for printing etc. Bear in mind that I have scanned thousands of slides and even the unadjusted pics are very good. You should be able to achieve this as well so hopefully it is just a case of settings

[wesjr]

Bez & Crofter,

Thanks for your help. In my adjusted scan provided, I used PS/ CS4 to clean it up. I always make a duplicate layer then do levels, curves, contrast and SH adjustments. On this photo I felt the need to us Neat Image for NR. My final step was to use FocalBlade a PS plugin to do a selective sharpening on the dandelion. Please use the zoom tool in PS to go to 200% and you will see either jpg compression artifacts or some other odd blue colors in the UR of image nothing seems to really fix this in PS: cloning or blurring the BG.

Bez I did another scan yesterday on another slide that looks terrific. Can't find the dandelion image right now. On the clean scan I went into curves and did a reset. Then when I did a histogram adjustment it looked totally better. The inital histogram was indicating sever underexposure and only half way to the right. All images before were the same very underexposed.

Bez I don't want my scans a little dark for fear of noise. Will do if its needed. Bez & Crofter: Do you do an auto exposure, histogram or tone/ curves adjustment after doing a preview scan before doing the actual scan. I do use the professional mode with Epson Scan.
____________________
Charlie Wesley
St. Augustine Beach, FL

[bez]

The best way to reduce noise is a better scanner 4990 is good; V700 slightly better; Nikon 9000 ditto, all combined with multi-sample which some third party software offers?
Turn Thumbnail off; when a tight-framed, zoomed preview is marqueed, hit the auto/reset, then mess about with histogram+brightness/contrast.
For colour you can go straight to balance RGB channels.

[wesjr]

Bez,

Thanks for the tips. I actually think the 4990 is a fine scanner. I'm not complaining. Not adverse to using Neat Image if it's needed. Thanks for the help...
_______________________
Charlie Wesley
St Augustine Beach, FL

[The Crofter]

Random example. This is a section from a watercolour that was scanned in six pieces, joined and then printed. Automatic adjustments were switched off so this is effectively raw. Apart from a slight increase in brightness the final image required no other adjustments and produced a near perfect print. Arguably I could have extracted a bit more detail by manipulating the settings during scan, but not much. The final print was on Hahnemuhle paper.

The point here is this is a good example of what the 4990 is capable of without "interference" from additional adjustments and also illustrates why I no longer worry about tweeking settings. It is possible that I have an extremely good machine and that there will probably be some variation between units. Did you buy your 4990 new or second hand ?.

watercolour 1.JPG (83.02 KiB) Viewed 2094 times

watercolour1 hist.JPG (14.44 KiB) Viewed 2088 times

[wesjr]

Crofter,

Thanks for your kind reply. Love your water color scan. It looks awesome. I now use the tones/ curves adjust tool with my scanner and I'm getting better results with much less noise.
I did buy my 4990 used from Amazon.com. It was really well cared for. The flatbed glass is spotless.
____________________
Charlie Wesley
St. Augustine Beach, FL

[The Crofter]

In fairness I think that the way I use my scanner is probably not strictly the best but it works for me. However I may be checking the output a bit more closely from now on

[lnbolch]

I have noticed that the EpsonScan histogram is a bit more forgiving than the one in Photoshop. If you just click auto, it looks like a bit of clipping in the shadows and substantial clipping in the highlights. However once, open in Photoshop, there seems to be a strong bias to protect highlight detail, with shadows normal, about the way I would usually want them.

Initially, I made the effort to move the markers at both ends of the scale to make sure I captured all the dynamic range, only to have to restore it in processing.



As you see, the EpsonScan histogram on the left looks like it is severely clipping the highlights, where the Photoshop histogram of the finished scan shows a flat-line indicating that only a few pixels are at full white. I use my eyes on the preview now and if it looks like the scan needs a bit of adjustment, then I do so. Otherwise, I just go with EpsonScan's guess, which seems to be pretty good in most cases.

Obviously, there is no 1:1 correlation between the two histograms. It is the one in Photoshop that counts, since that determines the image quality of the final published photograph.

[wesjr]

Larry,

Thanks for the reply. My EpsonScan histogram is very very limited in range also. Mine appears to clips the shadows while yours clips the highlights. Now I know it's OK in PS from your observations, I feel much better and can move forward.

Thanks so much....
____________________
Charlie Wesley
St Augustine Beach, FL

[lnbolch]

It will vary with the original image. The very long scale with the EpsonScan histogram gives some idea of the potential dynamic range of the scanner. I have never found any image that filled the whole scale, so far. Since I was scanning a negative, it took up relatively little room on the horizontal scale with this example.

[Kevgermany]

Sorry, have only just seen this thread.

The scanner is a fixed exposure device. That's why the initial scan never fills the histogram range. It doesn't give more (longer) exposure to darker material. All the scanner software does is take the scanned RGB values and amplify them. Which is why dark original films tend to scan very noisy, compared to what yuo'd expect from a darkroom print, where the pinter will incerase the exposure - and as there's no noise, this doesn't affect image quality.

Autoexposue just expands the input range to fill the output colour space. And therein lies a problem - if the output space from the scanner is wider than the space you use in photoshop, there can be some clipping. Similarly if the space output from Epson is smaller than the space you use in photoshop, it may look as if there's no clipping.

The left input scan in Larry's example will cause clipped output. Whether this is siginficant in the final image, I can't say. I'd guess not, based on his comments.

One other thing on the autoexposure. It's perfectly capable of clipping a single channel, and you'll see the black/white sliders sitting slightly inside the histogram ends when this happens, simliarly if the black/white sliders are outside the shown histogram after auto exposure, it's because one channel has a larger range than the other channels.

Manual histogram adjustments that ignore the 3 colour channels will frequently cause clipping in individual channels.

Noise - the histogram always shows dark to light from left to right. Negatives scanned as negatives are converted to positives before the histogram values are calculated. So scanning negs, where the darker, noisier parts of the film are the bright areas of the final image can bring up noise that you may not initially recognise as coming from the film. Another issue here is the maths involved. We're working with linear output from the scanner sensor, but an exponential curve for exposure. Net result is that the scanner has less numerical values to cover darker film areas, so resolves a lot less detail and is more likely to posterise the result. Add onto this the amplification needed to get shadow detail and the result is that most scanners can't properly resolve the darker areas of film, it just degenerates to a noisy, often posterised mess.

Slide films generally have a wider histgram than neg films - because the neg films have a wider exposure latitude. This can result in noise, particularly where the bulk of the scan histogram is to the right of the x axis. Be careful here, and always scan in 16bit mode. Banding, posterisation are also big issues resulting from the narrow density band of a typical neg film. Consumer oriented neg films seem to be worse, as they're designed with a higher exposure latitude than the pro films, so they cope bette with bad/non-existant exposure controls on point and shoot cameras.

If you really want to push the scanner, get a film target, calibrate the scanner and scan raw, applying the profile in PS. BUT - this prevents you from using ICE. Or apply the profile in Epson scan, which does allow ICE, but can lead to clipping if you're not careful.

Another approach is Vuescan, which will generate scanne pofiles if you have a taget, AND allow multi pass scans, so reducing the noise effects of dark film. Seems to do a good job as well, certainly on my microtek, where it's a lot bette than the supplied (and expensive) Silverfast. (Profiles are useless for negs, btw.). Order IT/8 tagets from Wolf Faust, http://www.coloraid.de.

Hope this helps.

[wesjr]

Kev,

Thanks for explanation. Now I know what's going on with the scanner. Your information will help me greatly.

It sounds like what I noticed when printing slides on Cibachrome. Be very picky on which slide you print. I will get some nice slide scans with my 4990 and other will not be worth the effort.

Thanks so much...
______________________
Charlie Wesley
St Augustine Beach, FL

[bez]

As Kev explains, in my experience multi-pass scanning produces noticeably better results, particularly with b&w negs. The Nikon 9000 scans are not only sharper (need less than half as much as the V700) but also show much better natural shadow detail, with just x2 multi-sample (it goes up x16!)