Seriously bad reflective Epson V850 scans

[donbga]

You nerds are making this difficult. Blur each patch and then read the values. That is of course assuming you are scanning properly. Just use the Epson software in Pro mode and set your white and black points properly.

Once that is done select a patch with the select tool of your choice and give it a Gaussian blur. Read the value 0 to 255. The use of layers is recommended.

Mark Nelson has a good explanation for this method in his Precision Digital Negatives software for the Mac. Worth every penny. Don't re-invent the wheel. Life is too short.

Don Bryant

[Doug Gray]

The best term for this seems to be "spatial crosstalk." I've found some discussion of this in sensor arrays and camera imaging systems where aperture, lens diffraction, and glare are the culprits. There are even specialized charts  for quantifying it. It's apparently long been known and pretty much ignored.

ISO 12641 specs the IT8 targets and usage. One problem with IT8 targets, is that they are for scanning traditional (chemical) photographs. More and more photos are not on film but digital. Arguably the original digital images never need scanning but as digital prints get older and the originals are lost, there will be a shift towards scanning inkjet prints. For those it's best to use targets made for the inkjet inks. This substantially reduces metameric shift errors since scanners are no where near L/I.

IEC 61966-8 provides lots of characterization methodology for scanners and is nearing two decades old. Section 13.1-13.4 focuses on scanner "Large area spatial crosstalk."
https://webstore.iec.ch/preview/info_iec61966-8%7Bed1.0%7Den.pdf

Running scannerreflfix.exe even improves the IT8 profile generating accuracy.

[nirpat89]

You nerds are making this difficult. Blur each patch and then read the values. That is of course assuming you are scanning properly. Just use the Epson software in Pro mode and set your white and black points properly.

Once that is done select a patch with the select tool of your choice and give it a Gaussian blur. Read the value 0 to 255. The use of layers is recommended.

Mark Nelson has a good explanation for this method in his Precision Digital Negatives software for the Mac. Worth every penny. Don't re-invent the wheel. Life is too short.


Don Bryant

Didn't realize PDN has solved all problems that were to be solved.

[donbga]

For what it costs the customer it's worth every cent. Mark has developed a very sophisticated package. Epson scanners have issues and the reflective glass isn't usually one of those unless the inside of the scanner needs to cleaner from out gassing.

Ik you want Lab values just pick up a spectro. Less expensive than the 850 scanner.

IT8 reflection targets for profiling the scanner isn't a bad investment either.
Mark's ebook provides an education on many levels for about the cost of a package of premiun inkjet paper.

[nirpat89]

For what it costs the customer it's worth every cent. Mark has developed a very sophisticated package. Epson scanners have issues and the reflective glass isn't usually one of those unless the inside of the scanner needs to cleaner from out gassing.

Ik you want Lab values just pick up a spectro. Less expensive than the 850 scanner.

IT8 reflection targets for profiling the scanner isn't a bad investment either.
Mark's ebook provides an education on many levels for about the cost of a package of premiun inkjet paper.

Hi, Don:

Not sure you are following the main thrust of Doug's work.  It is nothing to do with digital negatives, nor does it have to do with un-clean glass - Doug's scanner is brand new.  The phenomenon is related to the way a scanner gathers data from a given area which would be applicable, in general, to all scanners.  I joined in the discussion since I had seen a similar phenomenon while calibrating for digital negatives.

I do have a spectro now (ColorMunki Photo) which I plan to use in future.  The good thing about the scanner is being able to collect all the data in one shot.  Also have an it8 target.  I am fairly happy with the system I have developed for myself and the level of understanding on what is going on in making of digital negatives.  What I don't know, these forums have been excellent source of knowledge from experts like Doug Gray and others.  Specially if one knows the right question to ask.  I have nothing against the PDN system and it is not about the cost, it is about the satisfaction and fun of figuring things out yourself (that's the geek in me. ) 

:Niranjan.

[BrianToth]

It should be used only on 200 DPI scans from the Epson v800/v850 and is best used on RAW scan files. RAW files are the scan files made when you scan an IT8  target to run through I1Profiler. It works pretty well with images in ProPhoto and Adobe RGB as well but is slightly off because the native scanner gamma measures 1.7 so it will slightly overcorrect images in these colorspaces.

When you scan "RAW"… what software & settings are you using?

[Doug Gray]

When you scan "RAW"… what software & settings are you using?

Basic Epson scanner utility.

Using "No Color Correction" in the configuration menu. It works almost as good with ProPhoto RGB since the gamma of 1.8 is close to the 1.7 of the native scanner. Scanners don't have CFAs so RAW can be considered the same as native without doing anything else like color correction, sharpening, etc.

[Doug Gray]

I created a randomized set of 4mm patches using only white, green, yellow, black and blue colors in order to see the interactive effects of reflected light on the scanned patches. I ran the scanner fix program on the scanned target and generated histograms of each of the RGB channels on the scanned target, before and after reflected light correction. This shows the much tighter grouping of the RGB channels from removal of the reflected light. There remains some variation due to the white paper not producing exactly the same RGB values across the surface. The residual is surprisingly small though.

Also included are reduced images of the scanned patches as well as the calculated, reflected light which is a very blurry version of the patch set. Essentially, the fixed image is just the initial scan with the calculated reflected light subtracted.

The lighter values are affected by reflected light more than darker ones and the darkest ones are mostly noise, not luminance change from reflected light.

[BrianToth]

Basic Epson scanner utility.

Using "No Color Correction" in the configuration menu. It works almost as good with ProPhoto RGB since the gamma of 1.8 is close to the 1.7 of the native scanner. Scanners don't have CFAs so RAW can be considered the same as native without doing anything else like color correction, sharpening, etc.

I should have also asked: when you made your profile using Argyll, what parameters did you go with? I've found the XYZ cLUT options to produce the smallest Delta Es, but I'm not sure if it's just overfitting the target patches and might not be the best for general purpose work.

[Doug Gray]

I should have also asked: when you made your profile using Argyll, what parameters did you go with? I've found the XYZ cLUT options to produce the smallest Delta Es, but I'm not sure if it's just overfitting the target patches and might not be the best for general purpose work.

xt2ti3 -i -v infile scanner1
colprof -v [-bn] -ua -qh -D scanner9800.icm -O scanner9800.icm scanner1
colprof -v [-bn] -ax -ua -qh -D scanner9800m.icm -O scanner9800m.icm scanner1

I usually leave out the -bn option as the reverse table, b2a0, isn't used.

XYZ Luts are slightly better. Probably because the scanner, like a camera, uses RGB filters and XYZ is a somewhat better match.

I've been using a 957 patch file for scanner profiling and it does a pretty good job creating scanner profiles but I'm in the process of making a 1914 patch tables because the errors that remain are mostly in the darker, and saturated blues and cyans. And, for various reasons, I prefer using the LAB PCS luts rather than the XYZ PCS. These profiles work far better on prints than using the IT8 target profile which isn't very good for scanning inkjet prints. It also suffers from having a really low L* max. It's about 92 on the IT8 that came with the scanner. The IT8 target has colors that exceed my printer's gamut and vice versa so I really need separate profiles for scanning chem. process photos and inkjet prints. I'm not sure which is better for scanning other things.

I've also completed the code for scannerreflfix.exe so it works with different DIP resolutions. It also has options to embed a scanner profile and simulate Rel Col instead of Abs. Col. It simulates only boosting the L* since generally one wants to retain the tint of whatever is getting scanned. But I may add full white point Rel. Col. simulation later.  Normally, when scanning something it's desired to retain the actual possible off white of the scanned image. But not always. I'll post updated code shortly.

[David Good]

When you scan "RAW"… what software & settings are you using?

Vuescan has an option to scan to a "RAW" file. It's a tiff with all the image data so it can be re-developed without having to re-scan. I used to use this with large scanning projects. It's essentially a master file.

Dave