How best to convert paper-based workflow images to the web?

[Brad P]

This question has been posed and answered many times . . . but it is 2017.  How does one best accurately publish numerous calibrated monitor TIFF files to the internet?

I'm talking about exceptionally well monitor color calibrated TIFF files that print well. Think NEC or EIZO monitors and bright white (negligible OBA) thick art paper.  The answer probably is not still a simple Lightroom or Photoshop export, or the old Photoshop web action.  Maybe.  Some process or program to downsize corrections for the blueish, multivariate casts on something like the average of everyone else in the world's current monitors and do so with good image color and texture detail would be helpful. 

[Benny Profane]

Are you asking how to save an image so that it displays as well on a monitor 2000 miles away in somebody's office, or maybe somebody's IPad, or, of course, billions of phones around the world? Because, you know, even though it is 2017, it's still impossible. Just too many variables.

[Jim Kasson]

This question has been posed and answered many times . . . but it is 2017.  How does one best accurately publish numerous calibrated monitor TIFF files to the internet?

I'm talking about exceptionally well monitor color calibrated TIFF files that print well. Think NEC or EIZO monitors and bright white (negligible OBA) thick art paper.  The answer probably is not still a simple Lightroom or Photoshop export, or the old Photoshop web action.  Maybe.  Some process or program to downsize corrections for the blueish, multivariate casts on something like the average of everyone else in the world's current monitors and do so with good image color and texture detail would be helpful.

What is a "monitor color calibrated TIFF file"? Is is a file in some color space that looks good on your calibrated monitor? Or is it a file in which the RGB triplets are not colors, but colorant specifications for your monitor?

Jim

[Brad P]

What is a "monitor color calibrated TIFF file"? Is is a file in some color space that looks good on your calibrated monitor? Or is it a file in which the RGB triplets are not colors, but colorant specifications for your monitor?

Jim

Good question Jim.  For more background I work with large digital camera files in a Photoshop/Lightroom platform in 16 bit Prophoto RGB.  I use an NEC monitor set to a 120 cd/m2 luminance value with a D55 white point (I know, unconventional) and 2.2 gamma.  I properly calibrate my paper and print on an HP Z3200 printer.  I am satisfied with what I see and produce on screen and print.  I am uploading a number of images to the web and am not satisfied with the color, contrast and sharpness that results from a straight downsizing and sRGB JPG conversion in Photoshop and Lightroom when viewed on my iPad or other consumer viewing devices.  The result is noticeably bluer, the contrast is different and it's sharper in appearance.   

I get that it's impossible to downsize and convert files produced in my paper based work flow to sRGB and get an image that displays perfectly on everybody's individual consumer screens.  Consumer screens vary widely and typically are much brighter, oriented toward a D65 white point.  But it should be theoretically possible to target display values that are a hypothetical median of everyone's consumer monitors.   I am looking for some process/software other than what I have been using so that I can expediently do that well with a large number of files. It would be a bonus to have a process that doesn't tend to give an oversharpened appearance to already sharpened images as it downsizes. 

[Doug Gray]

Good question Jim.  For more background I work with large digital camera files in a Photoshop/Lightroom platform in 16 bit Prophoto RGB.  I use an NEC monitor set to a 120 cd/m2 luminance value with a D55 white point (I know, unconventional) and 2.2 gamma.  I properly calibrate my paper and print on an HP Z3200 printer.  I am satisfied with what I see and produce on screen and print.  I am uploading a number of images to the web and am not satisfied with the color, contrast and sharpness that results from a straight downsizing and sRGB JPG conversion in Photoshop and Lightroom when viewed on my iPad or other consumer viewing devices.  The result is noticeably bluer, the contrast is different and it's sharper in appearance.   

I get that it's impossible to downsize and convert files produced in my paper based work flow to sRGB and get an image that displays perfectly on everybody's individual consumer screens.  Consumer screens vary widely and typically are much brighter, oriented toward a D65 white point.  But it should be theoretically possible to target display values that are a hypothetical median of everyone's consumer monitors.   I am looking for some process/software other than what I have been using so that I can expediently do that well with a large number of files. It would be a bonus to have a process that doesn't tend to give an oversharpened appearance to already sharpened images as it downsizes.

Since you are used to a monitor white point of D55 images you view on other devices will generally appear bluer with a D65 white point. Further, many devices have white points even bluer than D65.

You can control the sharpness of the image you export by adding Gaussian blur or resizing using bicubic smoother, but there is no control over how another device will render an image. Mobile devices can change saturation, tone curve, etc. No way to know and no standards control this. Manufacturer's generally set things so that viewed images will be perceived as better on their devices than competitor's devices.

[Wayne Fox]

First I would submit that if you indeed have processed your files to create an "exceptionally well monitor color calibrated TIFF files" that achieves a perfect match in your particular environment, the only way to confirm that your files are "accurately published" would be to set up the monitors that are being published to next to your display and calibrate and profile those monitor to match yours.  But if I were to bring my display (NEC) which is also very well calibrated and profiled,  and set it up next to your monitor, the two images in all likelihood will not match.  Yes they would be close, but my display is setup based on my viewing and evaluating station, as I assume yours is, and thus there are probably slight differences enough to alter the appearance of the file slightly.

As to the overall issue of getting our images to look "accurate" or maybe a better thought is somewhat "close" to what we see, this has been discussed many times and most disagree with my philosophy, laying the problem upon all those hundreds of millions of people out there for not understanding color management and setting up their new computers accordingly ... basically saying it's their problem.

Yet the problem exists more because we are doing something with our computers that 99.99999% of the computers out there are not expected to do, and that is manipulate our display to try and mimic a paper print so as to more accurately predict the results when printing.  As such, each of us end up with a unique "one of a kind" setup, meaning our image might not even look correct on our fellow photographers display, and in some cases our displays are not that great for doing what the majority of people do with their computers.

So while there is some truth that we have no clue what all those displays out there are like, we do know one thing, the rest of the world has a display that is probably quite a bit different than ours, and their displays compared to each other are closer than our display compared to theirs.

There are two approaches ... forget about it, figure it's their problem and just go with it.  To be honest that probably doesn't work out as bad as one might think, because human vision is very adaptive, and even though the average display is probably quite a bit brighter, and in some cases such as mine quite a bit cooler, it still looks good.  Does it look identical?  No, but anytime that actually happens is coincidental, and even then no one really even knows when that happens because you can't compare them side by side.

Another approach is to decide to make some type of adjustment to hit what might be thought of as an "average".  My particular work display is around 53k and 115 cdm/2.  But the typical display as delivered by manufacturers is probably more around 6500 and at minimum 160 cdm/2 (and probably brighter than that).  I have 2 displays, so I set up one for printing, and the other more typical of the average manufacturers setting (sRGB emulation, 6500k, 160 cdm/2).  When creating a file that is to be viewed on such a display such as for the internet, I will usually view the exported jpeg/sRGB image on this display.  In my case, rather than worrying about a match I don't compare them to each other, I just take a look to make sure it looks OK.  Generally most look pretty good, but there are some where a little density and contrast adjustment improves them.  When I happen to be at a store with computers setup, I sometimes bring up some of my images on various Mac's and Windows machines just to try and make sure they look OK. 

One downside is my work might not look as good when a fellow photographer looks at it on their color managed display.  When I post an image to a website such as this or over at getDPI, i do keep that in mind, so I don't tweak the file even if it looks like it needs an adjustment on the other display, since I'm assuming most who frequent such forums have made some effort to calibrate their display.

To your original question as to a newer solution to get the image to be "accurate" on every display out there, despite it being 2017, nothing has changed.  Most are using a web browser that is not color managed and a display setup to manufacturers defaults, so to try and get a file to accurately publish is really no different that it was several years ago.

[Brad P]

Thanks for the thoughtful response, Wayne.  I see you have struggled with the same viscous emotions!   I suspect you're right, but I'm still hoping someone knows of some kind of new fangled automatic well thought out program I can run, or maybe a set of photoshop macros.

On my website, I'm less concerned with impressing fellow photographers as with people who order prints online and upon receipt are surprised by a different color caste and contrast values then their computer displays.  So unless someone points out something different, maybe I'll have to break down and buy a second screen for web proofing (or less optimally using the NEC with a new color value to flip between sRGB and Spectraview).  Fortunately, a new screen wouldn't be so expensive this time!!  If you have any thoughts on a middle of the road cheap screen that would work please share.
 

[BobShaw]

How does one best accurately publish numerous calibrated monitor TIFF files to the internet?

I use still use Aperture because it is the best programme available. Any of Lightroom, Capture One etc would do the same. You just export an sRGB in the size that you want using preset that are either there or you create. Select All > Export. I would never use Photoshop as it is cumbersome.

[Jim Kasson]

Good question Jim.  For more background I work with large digital camera files in a Photoshop/Lightroom platform in 16 bit Prophoto RGB.  I use an NEC monitor set to a 120 cd/m2 luminance value with a D55 white point (I know, unconventional) and 2.2 gamma.  I properly calibrate my paper and print on an HP Z3200 printer.  I am satisfied with what I see and produce on screen and print.  I am uploading a number of images to the web and am not satisfied with the color, contrast and sharpness that results from a straight downsizing and sRGB JPG conversion in Photoshop and Lightroom when viewed on my iPad or other consumer viewing devices.  The result is noticeably bluer, the contrast is different and it's sharper in appearance.   

I get that it's impossible to downsize and convert files produced in my paper based work flow to sRGB and get an image that displays perfectly on everybody's individual consumer screens.  Consumer screens vary widely and typically are much brighter, oriented toward a D65 white point.  But it should be theoretically possible to target display values that are a hypothetical median of everyone's consumer monitors.   I am looking for some process/software other than what I have been using so that I can expediently do that well with a large number of files. It would be a bonus to have a process that doesn't tend to give an oversharpened appearance to already sharpened images as it downsizes.

The SRGB images will appear bluer if displayed on a calibrated sRGB monitor since the sRGB white point is 6500K. Are you using a calibrated sRGB monitor? In addition, the gamut of your NEC monitor, if it's one of the higher end ones, is much larger than sRGB. Any colors that you see on your NEC monitor that are outside of the sRGB gamut will be mapped into that gamut. How they are mapped depends on how the conversion is made. Most conversion methods map the out of gamut colors to the envelope of the sRGB gamut. There are problems with that.

In addition, you may be using a set of methods that apply sharpening to your images before they are written out as sRGB files. You can check this by performing the conversions in Ps and writing them out as TIFFs. Are they too sharp? Then the culprit is probably the method that you use to reduce the resolution of the images. Try several Ps methods -- some have sharpening built in. Bilinear interpolation does not, but it may be too soft for your purposes.

Do the conversions in Ps and tell us what you see. Then maybe we can get Lr set up to do what you want.

Jim

[Brad P]

The SRGB images will appear bluer if displayed on a calibrated sRGB monitor since the sRGB white point is 6500K. Are you using a calibrated sRGB monitor? In addition, the gamut of your NEC monitor, if it's one of the higher end ones, is much larger than sRGB. Any colors that you see on your NEC monitor that are outside of the sRGB gamut will be mapped into that gamut. How they are mapped depends on how the conversion is made. Most conversion methods map the out of gamut colors to the envelope of the sRGB gamut. There are problems with that.

Thanks again Jim.  Interesting point to call out that makes me wonder if different software would map to different sRGB results. It makes perfect sense that the mapping of the gamut colors from 16 Bit ProPhoto RGB to 8 bit sRGB is at least some part of the problem because of the color compression and different ranges and that aspect of conversion may just be simply unsolvable.   My monitor is high end NEC and the Spectraview calibration settings are D55, Gamma 2.2, 120 cd/M2, contrast ratio 200:1, color gamut full (native).  I know different arguments can be made for different settings, and I'd be happy to take critiques on that, but I know most of them and don't think that's the root cause of what I'm struggling with in this post.

In addition, you may be using a set of methods that apply sharpening to your images before they are written out as sRGB files. You can check this by performing the conversions in Ps and writing them out as TIFFs. Are they too sharp? Then the culprit is probably the method that you use to reduce the resolution of the images. Try several Ps methods -- some have sharpening built in. Bilinear interpolation does not, but it may be too soft for your purposes.

Do the conversions in Ps and tell us what you see. Then maybe we can get Lr set up to do what you want.

You got me on this one, and this could be part of a problem in my workflow (at least vis a vis conversion to web images).  I heavily favor sharpening results that are more micro contrast adjustments and sharpening techniques other than traditional line based Lightroom-style sharpening and that is now an ingrained part of my workflow (i.e., halos aren't the problem but maybe micro contrast is).  This works well for print, but large downscaling may have difficulties (my files can be 2 GB).  I have to run now but will try some downsizing methods to see if that helps and it may well.  Thanks for that suggestion.

[Jim Kasson]

  I heavily favor sharpening results that are more micro contrast adjustments and sharpening techniques other than traditional line based Lightroom-style sharpening and that is now an ingrained part of my workflow (i.e., halos aren't the problem but maybe micro contrast is).  This works well for print, but large downscaling may have difficulties (my files can be 2 GB).  I have to run now but will try some downsizing methods to see if that helps and it may well.  Thanks for that suggestion.

At the risk of giving you TMI, here is a link to some of my reports on downsampling:

http://blog.kasson.com/?s=downsampling

Jim

[Benny Profane]

This was the ultimate prize promised by so many color management gurus when things got hot in the early stages of the craft: that the cost of final proofs could be eliminated, or severely reduced, when monitors could be calibrated in many different locations within a workflow. Proofs are expensive. Never happened. They tried, and are probably still trying, but, too many variables that require good old expensive manpower to manage. Just think of your monitor calibrated tiffs as your own, part of your own workflow, and leave it at that. Make a proof, hand it off, and say, hey, match that, please. Furgetabout the Internet.

[Ernst Dinkla]

Good question Jim.  For more background I work with large digital camera files in a Photoshop/Lightroom platform in 16 bit Prophoto RGB.  I use an NEC monitor set to a 120 cd/m2 luminance value with a D55 white point (I know, unconventional) and 2.2 gamma.  I properly calibrate my paper and print on an HP Z3200 printer.  I am satisfied with what I see and produce on screen and print.  I am uploading a number of images to the web and am not satisfied with the color, contrast and sharpness that results from a straight downsizing and sRGB JPG conversion in Photoshop and Lightroom when viewed on my iPad or other consumer viewing devices.  The result is noticeably bluer, the contrast is different and it's sharper in appearance.   

I get that it's impossible to downsize and convert files produced in my paper based work flow to sRGB and get an image that displays perfectly on everybody's individual consumer screens.  Consumer screens vary widely and typically are much brighter, oriented toward a D65 white point.  But it should be theoretically possible to target display values that are a hypothetical median of everyone's consumer monitors.   I am looking for some process/software other than what I have been using so that I can expediently do that well with a large number of files. It would be a bonus to have a process that doesn't tend to give an oversharpened appearance to already sharpened images as it downsizes.

My guess is that the iPad and several Android tablets are quite uniform in sRGB representation out of the box and a good target to aim at. At least compared to phones and consumer desktop systems. Even business systems outside the graphic, fashion etc industry are often not using CM and monitors vary widely on output. My son experienced this again when overhauling scientific plotting software for a major company here.

Next to color aliasing in downsampling is an important factor. Google "Bart van der Wolf aliasing" and I think you will be up to date on that aspect too.


Met vriendelijke groet, Ernst

http://www.pigment-print.com/spectralplots/spectrumviz_1.htm
November 2016 update, 700+ inkjet media white spectral plots

[Bart_van_der_Wolf]

I have to run now but will try some downsizing methods to see if that helps and it may well.

Hi Brad,

I've been looking into downsizing issues for more than a decade, and have some webpages that show the effects of different methods. They also give some tips on how to improve the standard Photoshop methods (by pre-blurring).

The webpage that is most popular can be found at its new location here. Some more (old, but still valid) examples can be found here.
Both pages are now hosted on a new server location.

The old webpage(s) from around 2004 have been updated last week to better handle modern web-browsers, and mobile devices. Some of the examples shown (i.e. the ImageMagick ones were updated from version 6 to 7) have also been updated. There will be info more coming (e.g. on edge halos and perhaps noise spectra), but I'm also working on a major overhaul of other old pages, on a new server, so it will take some time before that can be presented.

Lightroom's down-sampling is good, but Photoshop needs more work before good quality can be expected. Sharpening after downsampling can make artifacts more visible, but with Photoshop one has the opportunity to (or rather one should) use some additional tricks, like using a Blend-If sharpening layer to take the edge off of some of the issues. Also, plugins like Topaz Labs Detail offer great control over how the final image detail is rendered. A lot of that can be pre-recorded in an Action, so post resampling sharpening operations can be mostly automated.

I've also worked, here on LuLa, on creating a batch script file that uses various techniques provided by the ImageMagick applications to achieve even higher quality. That script does require to use the older version 6 of the ImageMagick suite, but I'm in the process of updating it so it can run on version 7. I'll let people know when that's finished. IM7 offers some possibilities that make it easier to further enhance the script, but we'll have to see if there is time for that.

Cheers,
Bart

[Jim Kasson]

Hi Brad,

I've been looking into downsizing issues for more than a decade, and have some webpages that show the effects of different methods. They also give some tips on how to improve the standard Photoshop methods (by pre-blurring).

The webpage that is most popular can be found at its new location here. Some more (old, but still valid) examples can be found here.
Both pages are now hosted on a new server location.

The old webpage(s) from around 2004 have been updated last week to better handle modern web-browsers, and mobile devices. Some of the examples shown (i.e. the ImageMagick ones were updated from version 6 to 7) have also been updated. There will be info more coming (e.g. on edge halos and perhaps noise spectra), but I'm also working on a major overhaul of other old pages, on a new server, so it will take some time before that can be presented.

Lightroom's down-sampling is good, but Photoshop needs more work before good quality can be expected. Sharpening after downsampling can make artifacts more visible, but with Photoshop one has the opportunity to (or rather one should) use some additional tricks, like using a Blend-If sharpening layer to take the edge of some of the issues. Also, plugins like Topaz Labs Detail offer great control over how the final image detail is rendered. A lot of that can be pre-recorded in an Action, so post resampling sharpening operations can be mostly automated.

I've also worked, here on LuLa, on creating a batch script file that uses various techniques provided by the ImageMagick applications to achieve even higher quality. That script does require to use the older version 6 of the ImageMagick suite, but I'm in the process of updating it so it can run on version 7. I'll let people know when that's finished. IM7 offers some possibilities that make it easier to further enhance the script, but we'll have to see if there is time for that.

Cheers,
Bart

Brad, listen to Bart when it comes to downsampling. He's the expert here. If you've read many of my blog posts on the subject, you'll have noticed that I am heavily in his debt.

Jim

[Brad P]

OK - I went away and actually read just about all the links, then tried some of it out.  Thank you Bart and Jim.  This has been very helpful and does clearly offers promise of improvement in image detail or texture (color as yet is unresolved as noted below).  Trying to sum up,

For Image Detail

* There is no single downsampling methodology that works best on all images.

* However, some methods are clearly better than others.  If one had to choose a single methodology to process, for example, 200 random images in a day, below are candidates of the best, ranked from most likely successful to least.
      - ImageMagick with Bart's free script.  This unfortunately will not work for me because I'm Mac based. If there's a way to get this to work on a Mac environment I'd like to (and not have to use cross platform software, which I could do but it's a pain).
      - ImageMagick with Lanczos filtering, anti-aliasing, and modest sharpening
      - Perfect Resize with modest sharpening
      - Lightroom with modest sharpening

What do you think of that list and order? 

But Color Remains a Big Problem . . .

What hasn't yet been solved for in this post is color casts.  I played around with an image (downsized TIFF attached) I'm working with in a stitched sunrise picture.  The seemingly correct WB value on my D55 screen is 9600K +36 tint.  I had both a polarizer and a 4 stop ND filter on at capture.  In this sunrise, we have slight shades of colors going through the sky and sand that's perfect for this illustration.  The colors change from a magenta orange into a light grayish green and a green/blue in the TIFF.  Using the Ernst Dinkla method, I simply emailed a Lightroom-produced JPG of this from my setup (5500K) to my iPad (6500K, confirmed by a Google search).  That resulted in an obvious blue cast on the iPad.  Then I bumped up the WB 1000K and emailed that, resulting in a diminished blue cast that's obviously still there.  Bumping the WB up to around 2500 gets close to correcting the blue at the top, but now the sand on the bottom has become greenish, so I need to adjust tint down. 

So a few things seem maybe to be going on here.  One maybe is that color casts result when Lightroom jams all the colors in a 16 bit ProPhoto TIFF into an 8 bit sRGB JPG.  Another issue is that the Lightroom JPG color cast is not be resolved by a simple proportional adjustment of WB and tint between the viewing devices.  It sure would be nice to find some kind of automated solution to this without having to view the images on separate monitors and make manual adjustments to each one.  Does anyone know of any more simple software or macro cure?  I suppose through trial and error I could try to write a photoshop macro for this, but it sure would be more convenient to find that someone already has!   

[Jim Kasson]

OK - I went away and actually read just about all the links, then tried some of it out.  Thank you Bart and Jim.  This has been very helpful and does clearly offers promise of improvement in image detail or texture (color as yet is unresolved as noted below).  Trying to sum up,

For Image Detail

* There is no single downsampling methodology that works best on all images.

* However, some methods are clearly better than others.  If one had to choose a single methodology to process, for example, 200 random images in a day, below are candidates of the best, ranked from most likely successful to least.

      - ImageMagick with Bart's free script.  This unfortunately will not work for me because I'm Mac based. If there's a way to get this to work on a Mac environment I'd like to (and not have to use cross platform software, which I could do but it's a pain).

      - ImageMagick with Lanczos filtering, anti-aliasing, and modest sharpening

      - Perfect Resize with modest sharpening

      - Lightroom with modest sharpening

What do you think of that list and order? 

Not bad.

But Color Remains a Big Problem . . .

What hasn't yet been solved for in this post is color casts.  I played around with an image (TIFF attached) in a stitched sunrise picture.  The seemingly correct WB value on my D55 screen is 9600K +36 tint.  I had both a polarizer and a 4 stop ND filter on at capture.  In this sunrise, we have slight shades of colors going through the sky that's perfect for this illustration.  The colors change from a magenta orange into a light grayish green and a green/blue in the TIFF.  Using the Ernst Dinkla method, I simply emailed a Lightroom-produced JPG of this from my setup (5500K) to my iPad (6500K, confirmed by a Google search).  That resulted in an obvious blue cast on the iPad.  Then I bumped up the WB 1000K and emailed that, resulting in a diminished blue cast that's obviously still there.  Bumping the WB up to around 2500 gets close to correcting the blue at the top, but now the sand on the bottom has become greenish, so I need to adjust tint down. 

So a few things seem maybe to be going on here.  One maybe is that color casts result when Lightroom jams all the colors in a 16 bit ProPhoto TIFF into an 8 bit sRGB JPG.  Another issue is that the Lightroom JPG color cast is not be resolved by a simple proportional adjustment of WB and tint between the viewing devices.  It sure would be nice to find some kind of automated solution to this without having to view the images on separate monitors and make manual adjustments to each one.  Does anyone know of any more simple software or macro cure?  I suppose through trial and error I could try to write a photoshop macro for this, but it sure would be more convenient to find that someone already has!   

If you want to find out what's going on, I think it's a big mistake to send the image to an uncalibrated device which you expect to serve as a reference.

How does your sRGB image look like on your monitor (or a different, but calibrated, monitor) when it's set to sRGB?

What's it look like on your monitor with its usual calibration when the image is converted to sRGB?  Don't save the file, just do the conversion in Ps.

If you're looking for software that's going to map a large gamut image to a small gamut color space with no human involvement and with results as good as what the most adept human can do, you're going to look forever.

Jim

[Brad P]

If you want to find out what's going on, I think it's a big mistake to send the image to an uncalibrated device which you expect to serve as a reference.

How does your sRGB image look like on your monitor (or a different, but calibrated, monitor) when it's set to sRGB?

What's it look like on your monitor with its usual calibration when the image is converted to sRGB?  Don't save the file, just do the conversion in Ps.

If you're looking for software that's going to map a large gamut image to a small gamut color space with no human involvement and with results as good as what the most adept human can do, you're going to look forever.

Jim

Yeah you are right, Jim. 

* When flipping the NEC monitor to sRGB mode, the image looked about right on the NEC monitor with the +1000K WB adjustment alone.  My iPad must be bluer than the reported factory specs.  That would argue for an easier maybe macro-able fix than I thought I was facing at least to the sRGB color space which would be an improvement.

* Very different and interesting results when simply converting the image from ProPhoto RGB to sRGB.  Eyesight-wise, the luminosity and yellows in particular pop very noticibly.  Doing a few color sampler tests, on this particular image anyway the reds stay pretty constant with a slight bump up, but the blues increase quite a bit, and the greens even more. 

It all still leaves me wondering a bit about what the right target consumer display color targets would be.  These are not trivial differences.  I'm not going to give up yet on an automation.  This stuff seems like some kind of automation can at least get one a whole lot closer, and maybe the easiest part of that is to bump up WB to 6500 or so.  Though it's looking like images may need to be at least hand graded and adjusted at some point, maybe as Wayne suggested with two monitors side by side.

[Brad P]

Update on the remaining color issues:

Searching high and low for answers, I ran across this interesting article outlining fairly recent advances in web browsers and viewing devices.  Reading this and considering the above, it suggests to me the following workflow.  Any comments welcome.

http://www.color-management-guide.com/web-browser-color-management.html

* Continue my normal paper based workflow with ProPhoto RGB and color management for that.  This ain't changing anyway because print is mission critical. 

* For publishing to the web, downsize as I summarized others' comments above for image detail. 

* Increase the image's WB 1000K to compensate for the difference between my monitor's calibration and an assumed 6500K average viewing device.  (Others reading this should do the math based on their monitor's calibration).

* Convert the file to Adobe RGB, and embed that profile in the image.

This has advantages and disadvantages.   The advantages are that increasingly browsers and viewing devices are trending towards supporting a viewing environment at least approaching an Adobe RGB visual gamut.   Compressing ProPhoto RGB to Adobe RGB does less violence to the out of gamut colors than converting to sRGB, particularly in green space, and Adobe RGB better captures the color variations one would see from an ink jet print.  The disadvantage is that those using older monitors and browsers (including most non-iOS tablets) might be viewing what becomes an entirely non color managed file, which otherwise could have been improved by saving to sRGB instead.   So this workflow distinctly favors those who care about viewing device color or otherwise update their viewing hardware, and disfavors those who don't.   I think I'd rather take that approach rather than converting to sRGB and putting out a compromise. 

[GrahamBy]

Is distributing digital images for people to look at on uncalibrated monitors so different to distributing prints for people to look at under non-standard lighting?