Pages: [1] 2   Go Down

Author Topic: Seriously bad reflective Epson V850 scans  (Read 2189 times)

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1350
Seriously bad reflective Epson V850 scans
« on: May 20, 2018, 03:53:32 PM »

I see as much as a L* difference of 10 for highlights depending on what colors are around, and how close they are to the highlights. It is clearly due to light, reflecting off the paper and back onto the two frosted, LED light sources which then illuminates that portion of the paper which is being scanned.

Given the magnitude of a potential error, the visual effects are usually subtle but result in a slightly off copy when using the scanner for critical reproduction purposes.

Doing this will demonstrate the problem:

1. Create an image with 2, 4" by 4" black squares.
2. On the first insert a centered, 0.1" white square.
3. On the second insert a centered 3.5" by 3.5" white square.
4. Print it.

Scan them with a profile that doesn't clip the whites and examine the L* of the white patches in the center of each square. Prepare to see something like values of 89 and 99, or possibly higher levels like 94 and 100, if the whites are clipping above 100. Those are shocking differences.

Two big issues. Clipping of near whites, which is highly likely when scanning papers with high white L* over 96 with subtle high key tones. This can be mitigated by setting the "Scale" factor in Argyll's colprof. But the other issue is inability to colorimetrically replicate an original. Arguably, replication by colorimetric means is going to introduce large errors if the profile is generated from patch sets with different spectra than the scanned object but this just adds error. And it's a lot!

It doesn't show up in the normal dE scan tests because dE is reported self referentially. The scanned color of IT8 or any other patch set has a fixed set of patch locations so the impact of adjacent patches is baked into the profile and results in "good" readings on the same patches as those were what the profile was created from.

One positive note: This does not apply to scanning film. It's purely an effect of bright areas of the paper reflecting light back down on the horizontal frosted illumination strips which then increases the light seen by the scanned portion. Negligible effect scanning transparencies.
Logged

kers

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1810
    • Pieter Kers
Re: Seriously bad reflective Epson V850 scans
« Reply #1 on: May 20, 2018, 05:15:41 PM »

Is it possible it also has to do with the glasscover of the scanner ?
That was my basic problem always and the dust being trapped between...
Always wanted to cut a hole in the glass but then you introduce new problems
Logged
Pieter Kers
www.beeld.nu

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1350
Re: Seriously bad reflective Epson V850 scans
« Reply #2 on: May 20, 2018, 05:24:51 PM »

Is it possible it also has to do with the glasscover of the scanner ?
That was my basic problem always and the dust being trapped between...
Always wanted to cut a hole in the glass but then you introduce new problems

Nope. The amount of light difference in the two examples I gave is about 25% (in linear space such as nits, lux, etc.). It's simply huge. But it doesn't affect black patches very much at all because a patch with L* of 4, illuminated by 25% more light, only goes up to 5.  This is entirely consistent with light from the illuminated areas of the paper reflecting back onto the frosted light sources, which are quite close to the paper, and effectively increasing the illumination the paper gets at the scan location.

This is a problem creating scanner profiles as the scanned RGB values are a function of the patch color as well as the colors of patches within about 1/3 inch of it horizontally and about 1/2 inch vertically. After that distance the errors drop by over 90%. This, it turns out, offers a mathematical way to compensate for the scanner's extra reflected light. Since patch coordinates are regular, the scanned values, converted to linear XYZ, can be multiplied by the fraction that represents their contribution which is a fixed function of geometry. Then the patches' XYZ values can be proportionally increased by the total increased illumination the patch is receiving to arrive at an XYZ representing the patches' color were it to be illuminated properly. This processing can be done quite automatically and should produce much better profiles.

I am in the process of doing this now.



However, while if fixes most of the errors that go into making profiles, it does not fix errors from subsequent scans of other images.

But something similar can be done. A 2D map can be made where each point on the map represents the relative fraction of increased illuminance from all the colors that are close enough to significantly reflect back to frosted LED bars and on to the scanned point. This could easily be programmed in C or C++ to read a tiff file, process it, and write out a corrected tif file.

The somewhat, but not too hard part, is determining the contribution function.
« Last Edit: May 20, 2018, 05:49:50 PM by Doug Gray »
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1350
Re: Seriously bad reflective Epson V850 scans
« Reply #3 on: May 23, 2018, 02:10:57 AM »

I made a test gradient from L*90 to L*100 to test the impact of a surrounding color on the accuracy of the V850 scanner. Each gradient was inserted into a white and black surround then printed using Relative Colorimetric so that the full gradient appears on the print and looks the same, aside from tint, as the paper. Prints were made using Costco Glossy (paper white L*95) and a Baryta luster (Paper White L*98). Scans were made using the profile generated from the provided IT8 chart using I1Profiler.  Each the two gradients from both papers were converted to ProPhoto and attached below. They show the impact of a black surround v a white surround.

This thread relates to a problem scanning a stamp, which was inside a white surround and this may be the proximate cause. Note that because perception of brightness is nonlinear, whites in an average scene surround (presumably around L*50) would see 80% of the drop show here by the black surround.

http://forum.luminous-landscape.com/index.php?topic=72000.msg571978#msg571978


Logged

nirpat89

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 264
    • Photography by Niranjan Patel
Re: Seriously bad reflective Epson V850 scans
« Reply #4 on: May 23, 2018, 12:04:30 PM »

In my experience with using a scanner (mine is an Epson Perfection 3200) to measure densities to calibrate my digital negatives for alternative printing processes, I have encountered similar spatial dependency of data.  For example, the density on the border frequently seems to be greater than a patch (of the same exposure) in the middle.  I chalked it up to some form of chemical/kinetics based phenomenon during develop/fixing processes.  Based on your data, it could well be a scanning artifact. 

It would seem to me that the scanner is not a means to do precision measurement work like a spectrophotometer.  For example, I have noticed that where you put the print on the glass can also affect the readings.  If I place the print, measure and rotate it around and measure again, there could be as much as 5-10% variation in the readings.  So for comparison purposes I try to place the print the same way each time.

:Niranjan.
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1350
Re: Seriously bad reflective Epson V850 scans
« Reply #5 on: May 23, 2018, 02:05:44 PM »

In my experience with using a scanner (mine is an Epson Perfection 3200) to measure densities to calibrate my digital negatives for alternative printing processes, I have encountered similar spatial dependency of data.  For example, the density on the border frequently seems to be greater than a patch (of the same exposure) in the middle.  I chalked it up to some form of chemical/kinetics based phenomenon during develop/fixing processes.  Based on your data, it could well be a scanning artifact.
I'd bet it's a similar effect to what I'm seeing. When I first looked at the way prints were illuminated by the scanner my first expectation was that there would be variations in luminance across horizontal scans. LEDs are notoriously variable when trying to match withing 10% let alone 1% or 2%. But much to my surprise, there was almost no horizontal variation. It was under 1% except at the scan edges. Almost certainly the scanners are individually calibrated and horizontal positional gain is written to nvrom during manufacture. I was happy to see this.
Quote

It would seem to me that the scanner is not a means to do precision measurement work like a spectrophotometer.  For example, I have noticed that where you put the print on the glass can also affect the readings.  If I place the print, measure and rotate it around and measure again, there could be as much as 5-10% variation in the readings.  So for comparison purposes I try to place the print the same way each time.

:Niranjan.

A spectro is very precise, and provides highly repeatable measurements. Especially the ones that use white LEDs as an illuminant which are more stable than tungsten lamps which have thermal drift time changes. Even so, the latter is consistent within 1% or so and the former perhaps 0.1%.

The problem with spectros is that they can't measure very small areas. They effectively average the reflected light over about 5 mm^2.  A scanner, like a camera, can measure luminance over areas of hundred um*2 to better than 1% quite well. But they can't measure spectral info. But they do a reasonable job of measuring colorimetric info. Especially when measuring CYMK printed images or RGB ones from a display (for a camera). This is because the spectral info can be modeled by a linear combination of the CYM inks or RGB colors. Problems arise when scanning or photographing things with more spectral diversity such as colorcheckers, especially the Colorchecker SG. There, the critical factor is how well the scanner/camera meets the L/I criteria.

While I was initially concerned about how much light is reflected by the glass or other structural components unrelated to the paper's actual reflectance, it turns out this is pretty small. No more than about .2%.  It's so low that if I scan with nothing on the scanner and the lid open with the room lights off I can see residuals on the glass like fingerprints, or even ink patterns from scans of prior prints. To mitigate this I've started baking my prints at 160F for half an hour after printing to reduce/eliminate the small amount of glycol that is mixed with water in the inks. Letting it dry at room temp. even overnight was not sufficient. Amazing effect actually.

In looking at the actual reflected light ratios, a white sheet of Costco paper measures 1.20 times higher than a white patch that is 3mm square inside an otherwise black inked sheet. For the Baryta paper, the ratio is a bit higher at 1.22. This is consistent with the Baryta spectral measurements which indicate it reflects 10% more light than the Costco paper.  1.22 = 1 + (10% of .2).  Note that the ".2" in that equation is the portion of additional light reflected by white paper's surface back onto the frosted LED covers, then back to the paper.

A validation of the hypothesis is doing the same experiment when the "white" is decreased by inking so the reflectance is half that of the uninked paper. The same experiment yields only a 1.10 difference on the Costco papers. This is consistent with halving the light reflected from the LED covers while the full light directly from the LEDs is still hitting the scanned paper surface.

Fortunately, this turns out to be correctable in two phases. The first is the correct each scanned patch by subtracting the contributions of patches around it. This basically involves multiplying the spectral data of each patch by it's positional reflectance contribution (from about 1.2% to .05% depending on position and doing this for the 24 surrounding patches. This goes out to about 20mm around the 6mm patches generated for the iSis and, while there is some impact beyond this, it's minimal and this approach captures over 90% of the scanner, reflected light, additive error. I've got the numbers for this and am incorporating it into the process that creates the scan data for Argyll.

The second phase involves reading the RGB values of a scan's pixels, calculating the XYZ values, and doing a similar subtraction. While this rapidly expands into some really large calculations it can be simplified by resizing the pixels used for the adjustment by 10 to 1 or more.
« Last Edit: May 23, 2018, 02:09:40 PM by Doug Gray »
Logged

BrianToth

  • Newbie
  • *
  • Offline Offline
  • Posts: 42
Re: Seriously bad reflective Epson V850 scans
« Reply #6 on: May 23, 2018, 11:42:41 PM »

I don't mean to hijack your thread, I've been finding your process of digging into the problem interesting and curious to see where it goes.

Not to oversimplify, but do you think this is a common issue with LEDs?  I still haven't overcome the color casts when trying to profile my v800 (http://forum.luminous-landscape.com/index.php?topic=124117.0) but my observations have been as follows, for what it's worth to you:

•  Profiling with any IT8 reflective target leads to overall bad blue casts. 
•  Profiling with a ColorChecker 24 or ColorChecker SG are a bit redder, to be unscientific, but are better overall with less extreme color shifts. Still not great.
•  Even the canned Epson profiles in Epson Scan, or SilverFast yield a blueish cast. 
•  I've tried scanning with a black background to cut down on errant reflections, but it didn't help. I'm primarily scanning glossy photos, which again might be related… I don't have any matte photos to try right now.
•  I printed a custom target using Argyll with 962 patches on Canon Glossy II and that profile does a really good job of reproducing my prints on Canon Glossy II paper or similar. But it completely fails with any of my real photos. So for the moment I've been getting by with Argyll's XYZ cLut.

I was getting the impression that the matte targets were reflecting less of the LED light than the glossy IT8s and that might be why they were working better. I was tempted to try to find an old CFL scanner to see if it gave me any better results because so far I've been disappointed. I have an old v300 that's also LED and has the same issues.

I don't see anyone else talking about issues like this, but it's very easy to reproduce by scanning and sampling prints of ColorCheckers. I thought I might be alone. :)

Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1350
Re: Seriously bad reflective Epson V850 scans
« Reply #7 on: May 24, 2018, 01:36:22 AM »

You have some interesting results. My main goal is to be able to scan prints, not photos, and expect that profiles that work well in one will not be as good in the other. The real purpose of trying to get good colorimetric scans of prints is that it will provide a mechanism to locate, and quantitatively measure printer smoothness anomalies that don't show up from limited patch sets.

So my focus currently is to characterize, and compensate for how scanner colors are altered by the surrounding patch colors via reflection. The problem isn't the LEDs but the reflected light that hits the frosted LED coverings and bounces back. However, this is largely something that distorts higher luminance patches. I've also found that there is considerable noise in the lower luminance patches and much of the dE variation I see occurs there. I'm not sure what the cause is but I'm going to look at a histogram of the these patches. One thing's for sure, a bright pixel on a black background has far more impact that a black pixel on a white background. But drilling into that is going to wait until I get the reflectance compensation working. I currently have a 2D matrix with factors for the luminance impact of adjacent patches.

To show a bit of the impact of this, I have a scanner profile made from the standard iSis 957 patch set and when I scan the patch set and measure the dE00 of the 600 higher luminance patches, the average dE00 is about .51. OTOH, if I measure the same patches but position scrambled, the average dE00 goes up to 1.2. This is because the scanner profile was made with the first patch arrangement. The same effect can be seen with printer profiles due to slight variations in printer patch color but almost all of the scanner differences is due to the reflectance problem.

One thing I've noticed is that it is critical to good profile behavior in low luminance patches that the glass is exceedingly clean. I've seen patterning from residual glycol from the prints left on the glass.
« Last Edit: May 24, 2018, 01:55:19 AM by Doug Gray »
Logged

BrianToth

  • Newbie
  • *
  • Offline Offline
  • Posts: 42
Re: Seriously bad reflective Epson V850 scans
« Reply #8 on: May 26, 2018, 02:17:42 PM »

I happened to be printing out some grey patches for a different test of mine I thought I'd see if I could replicate your experience on my scanner. I printed them with both a white and black background and measured the centers and edges of each patch to see the effect of the neighboring color. Not as thorough as your tests of course, but seems to suggest it's a factor here also. (I attached my partial findings.)

From your testing, do you think this is a bigger issue for targets like an IT8 or Hutch that have tiny patches crammed right next to each other vs other targets? I'm just curious if you think the issue can be mitigated by a different method.  Both of the ColorCheckers for example offer bigger patch areas, matte black areas between each patch, slightly recessed patches, and a thick background, but they have different spectral properties than a lab photo.  If printing your own patches, could you print larger patches and make sure to measure as much in center as possible? (Maybe your automated equipment doesn't offer you that flexibility.)

For example, I have noticed that where you put the print on the glass can also affect the readings.  If I place the print, measure and rotate it around and measure again, there could be as much as 5-10% variation in the readings.  So for comparison purposes I try to place the print the same way each time.

For what it's worth, I was curious about how much that could play a role in my regular scanning so I did a test of scanning an IT8 four times: each way rotated differently and in a different area of the glass. I then took the four images into Photoshop, straightened and rotated them, and did a "difference" blend between them and saw no difference between the patch values… so at least that seems like a good thing in my case. (Assuming my test was valid.)
Logged

Stephen G

  • Full Member
  • ***
  • Offline Offline
  • Posts: 126
Re: Seriously bad reflective Epson V850 scans
« Reply #9 on: May 27, 2018, 04:30:53 AM »

I've been following this thread with interest, as I rely on my V700 to scan artworks. The V700 has a CCFL lightsource. I ran the test that Doug suggested in his first post to this thread. See attached. The whites read at about L90 and L99 for the small and large white squares respectively, so I'm getting the same effect.

Process: printed the squares on Epson Enhanced Matte. Scanned to raw 16-bit linear tiff through Vuescan. Assigned my profile, which is based on a ColorcheckerDCSG scanned in the same way. Converted to working space then dropped to sRGB and 8-bit for web.

I'm now wondering how much of a role this effect has played in my work with this scanner, over the past seven years that I've used it. The upper tones of watercolours - washes that are very close to paper white in brightness, and paper white itself - have always had a tendency to clip or come close to clipping. I'm also wondering if there's anything I can do to counteract the effect. Perhaps that will come of your work here.

If you think it will help I will run further tests to contribute to the data. I'm no coder/mathematician so I can't really help with that side of things.
Logged

nirpat89

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 264
    • Photography by Niranjan Patel
Re: Seriously bad reflective Epson V850 scans
« Reply #10 on: May 27, 2018, 09:44:20 AM »

For what it's worth, I was curious about how much that could play a role in my regular scanning so I did a test of scanning an IT8 four times: each way rotated differently and in a different area of the glass. I then took the four images into Photoshop, straightened and rotated them, and did a "difference" blend between them and saw no difference between the patch values… so at least that seems like a good thing in my case. (Assuming my test was valid.)

Could be that my scanner is unique to this problem.  It is pretty old.  It is also based on CCFL light source  - perhaps those are not uniform across the width as LEDs might be.  Or may be I need a new tube which I have not changed since buying the scanner may be 10-12 years ago.

:Niranjan.
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1350
Re: Seriously bad reflective Epson V850 scans
« Reply #11 on: May 27, 2018, 05:50:09 PM »

For those that have issues with high luminance blowout, which can occur with very "white" papers and find detail near the whites, one crude approach is to use unmask sharpening in Photoshop with a very wide radius. At 200 PPI, a radius of about 80 and fraction of .1 or so does a good job of removing the reflected light luminous increases. Then, using curves, decrease the tone curve by about 10%. It should be applied to the raw tif file, then the scanner profile s/b assigned. However, it doesn't work in middle tone neutrals because the reflected light impact isn't linear. An example:

Taking a case where a white patch has a reflectance of .80, and 10% of the light (assuming a perfectly reflecting paper white) is reflected from the frosted LED covers, the total light scanned will be .80 + .10*.80*.80 or .864.  But if the paper had a reflectance of .40 (50% less than paper white), the formula is .40 + .10*.40*.40 or .416  and so the scanned value is too low.

The reason for the .80*.80, etc. because the additional reflected light is first attenuated by the paper then, when reflected back to the paper is attenuated again by the paper's density. There may well be a mechanism in Photoshop using multiple layers. First an unsharp mask, then a multiplicative layer that attenuates the unsharp mask. I'm not a Photoshop expert enough to pursue this.

I believe the best way to handle this is a standalone program that reads a tif file, and applies a transform adjusting for these effects. The program needs nothing more than a RAW scanned tif file but would be specific to a particular scanner model. That's my goal and do far I haven't encountered anything that suggests it's not doable. I'm checking the model in Matlab but that's not a practical general solution. Should have C++ code in a few days. Should be something simple like "fixv850scan.exe inTifFile, outTifFile"
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1350
Re: Seriously bad reflective Epson V850 scans
« Reply #12 on: May 30, 2018, 05:06:26 PM »

Yay. I have a fully working set of Matlab functions that implement reflected light correction. I'm going to do some tests on this then on to a C++ stand alone program that will fix up any scan from an Epson V800 or V850.
Logged

nirpat89

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 264
    • Photography by Niranjan Patel
Re: Seriously bad reflective Epson V850 scans
« Reply #13 on: May 31, 2018, 08:13:06 AM »

Yay. I have a fully working set of Matlab functions that implement reflected light correction. I'm going to do some tests on this then on to a C++ stand alone program that will fix up any scan from an Epson V800 or V850.

Doug:

If I were to design a step-wedge that minimizes cross-contamination between adjacent blocks, what would you recommend for a good layout scheme?  Individual blocks surrounded by a black border? How big?

:Niranjan.
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1350
Re: Seriously bad reflective Epson V850 scans
« Reply #14 on: May 31, 2018, 01:51:07 PM »

Doug:

If I were to design a step-wedge that minimizes cross-contamination between adjacent blocks, what would you recommend for a good layout scheme?  Individual blocks surrounded by a black border? How big?

:Niranjan.

If the goal is minimizing non-linear effects from reflected light (cross contamination) then making small patches of 3mm separated by about 6mm or more with black centered on a 40mm wide strip would work pretty well. However, uncorrected profiles will distort measurements since they are made with contaminated patches. Using the raw  RGB values and correcting only for a constant gamma works quite well. The native gamma of the V850 is very close to 1.7.

OTOH, if the goal is to prevent whites and near whites from saturating on a high L* paper, you can get by with a smaller black surround and use larger, adjacent rectangular patches but the lighter ones will cross contaminate and even contaminate their own readings. But this will largely eliminate white and near white saturation on high L* paper.

For instance even a single white patch 6mm on a side around a large black will still have luminance increased a few percent in linear XYZ space. Cutting the size down to 3mm on a side reduces the increase to about .5% which is only about .2 dE76 or .1 dE00.
Logged

nirpat89

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 264
    • Photography by Niranjan Patel
Re: Seriously bad reflective Epson V850 scans
« Reply #15 on: June 02, 2018, 09:37:46 AM »

If the goal is minimizing non-linear effects from reflected light (cross contamination) then making small patches of 3mm separated by about 6mm or more with black centered on a 40mm wide strip would work pretty well. However, uncorrected profiles will distort measurements since they are made with contaminated patches. Using the raw  RGB values and correcting only for a constant gamma works quite well. The native gamma of the V850 is very close to 1.7.

OTOH, if the goal is to prevent whites and near whites from saturating on a high L* paper, you can get by with a smaller black surround and use larger, adjacent rectangular patches but the lighter ones will cross contaminate and even contaminate their own readings. But this will largely eliminate white and near white saturation on high L* paper.

For instance even a single white patch 6mm on a side around a large black will still have luminance increased a few percent in linear XYZ space. Cutting the size down to 3mm on a side reduces the increase to about .5% which is only about .2 dE76 or .1 dE00.

Thanks Doug for the recommendations.  My goal was to get better data from the scanner for use as a pseudo-densitometer to characterize an alternative printing process using a digital negative.  My typical step-wedge is a bunch of blocks separated by a thin black border, arranged in a square layout.  I made a new step-wedge by simply shrinking the blocks and filling the space with black, so now they are separated by about 7 mm between them all around.  (Attachment 1) 

To see what that did to the measurements, I printed both side-by-side on a glossy paper and scanned them.  They were then taken to Photoshop, de-staurated and their RGB measured using identical size probe (rectangular marquee tool) and mean values read form the histogram data.  The result shows a good bit of difference between the old and the new design.  The new design gave consistently darker numbers than the old design, with absolute amount ranging from 0.56 to 4.39.  Calculated on percentage basis, they ranged from 0.24% to 5.14%.  (Attachment 2)

Based on these results, looks like I would have to start using the new layout in the future... 

:Niranjan.
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1350
Re: Seriously bad reflective Epson V850 scans
« Reply #16 on: June 03, 2018, 01:30:31 AM »

Thanks Doug for the recommendations.  My goal was to get better data from the scanner for use as a pseudo-densitometer to characterize an alternative printing process using a digital negative.  My typical step-wedge is a bunch of blocks separated by a thin black border, arranged in a square layout.  I made a new step-wedge by simply shrinking the blocks and filling the space with black, so now they are separated by about 7 mm between them all around.  (Attachment 1) 

To see what that did to the measurements, I printed both side-by-side on a glossy paper and scanned them.  They were then taken to Photoshop, de-staurated and their RGB measured using identical size probe (rectangular marquee tool) and mean values read form the histogram data.  The result shows a good bit of difference between the old and the new design.  The new design gave consistently darker numbers than the old design, with absolute amount ranging from 0.56 to 4.39.  Calculated on percentage basis, they ranged from 0.24% to 5.14%.  (Attachment 2)

Based on these results, looks like I would have to start using the new layout in the future... 

:Niranjan.

I hate to tell you this but you need both a larger black surround and reduced area on the patches. They are still pretty large and you are getting significant boost in brightness on the bright patches because the area is still quite large. If you reduce the patches to 4mm on a side which will also expand the black surround you will likely see an additional 5% reduction in lightness in linear space (Y in XYZ notation) on the white patches.

« Last Edit: June 03, 2018, 01:38:48 AM by Doug Gray »
Logged

nirpat89

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 264
    • Photography by Niranjan Patel
Re: Seriously bad reflective Epson V850 scans
« Reply #17 on: June 03, 2018, 12:19:26 PM »

I hate to tell you this but you need both a larger black surround and reduced area on the patches. They are still pretty large and you are getting significant boost in brightness on the bright patches because the area is still quite large. If you reduce the patches to 4mm on a side which will also expand the black surround you will likely see an additional 5% reduction in lightness in linear space (Y in XYZ notation) on the white patches.
Understood.  The thick border is for reducing "cross-contamination," while the smaller block size is for reducing "self-contamination." 

(Your work, by the way, would be worth a lot of money to a scanner company.)

:Niranjan.
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1350
Re: Seriously bad reflective Epson V850 scans
« Reply #18 on: June 04, 2018, 02:55:57 PM »

(Your work, by the way, would be worth a lot of money to a scanner company.)

:Niranjan.

I don't think scanner companies would be much interested. What they have chosen to do is a tradeoff between scanning flat sheets and things that aren't as flat and rise above the glass a bit. Correcting for reflection would negatively impact scans of non-flat objects and the visual impact of flat sheet scans is minimal.

However, it would be a good subject for someone to do an academic paper on. Preliminary searches show the effect is not recognized let alone accounted for.
Logged

Doug Gray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1350
Re: Seriously bad reflective Epson V850 scans
« Reply #19 on: June 12, 2018, 04:08:56 PM »

I'm providing the fixup program in an attached zip file with executable and needed dlls for reading and writing tif files. It's used like this:

scannerreflfix file1.tif file1_fixed.tif

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. However, that error is much smaller than the reflectance error itself.

Profiles should be created using the reflection corrected tiff file whether an IT8 or homemade. I made a scanner profile targeting my printer by using a printed, standard printer target but scanning it then feeding the results, together with the iSis scanned lab values into Argyll's software. This produced quite a good profile with an average dE76  under 1.0 when scanning printed colors of reflection corrected scans.

Optimal workflow is to always scan to RAW then run scannerreflfix and attach the scanner profile to the result. Then convert to Adobe RGB or ProPhoto if the gamut extends beyond Adobe RGB.

The attached images show the scanned file and the reflected light corrected file, as converted by the scanner profile into Adobe RGB. Also an image that shows the magnitude of the reflected light distribution over the scan. You can see how the reflected light increases over the brighter areas.

Here's the front summary of the results comparing dEs with high L* neutrals surrounded by black or white as well as the 24 Colorchecker colors, all of which were printed surrounded by white.

_____

This file contains measured results from neutral patches and the set of ColorChecker colors. The first part are high key neutral patches, printer RelCol, from L*=82 to L*=100 in steps of 2 totaling 10 patches. These patches are printed surrounded by black, then whit,e areas. This represents the two, reflected light extremes. The patches with the white surround are boosted about 20% by all, near reflected light while the patches surrounded by black are boosted by light from the patch itself, which is about 3% for a 6mm patch. These are followed by the 24 colors in a ColorChecker, printed AbsCol, each surrounded by white.

The largest differences are between high L* colors that are surrounded by by black v white. The average difference dE before reflection correction is 6.07 dE, dropping to 0.83 dE after correction.

Uncorrected image.
The average deltaE's for these three groups:
Neutrals (black surround):     mean dE: 1.66
Neutrals (black surround):     mean dE: 4.64
ColorChecker (white surround)  mean dE: 3.33

After image corrected for reflected light.
The average deltaE's for these three groups:
Neutrals (black surround):     mean dE: 0.41
Neutrals (black surround):     mean dE: 0.80
ColorChecker (white surround)  mean dE: 0.65

« Last Edit: June 12, 2018, 04:12:21 PM by Doug Gray »
Logged
Pages: [1] 2   Go Up