Looking for feedback from people have experience printing

[chrisgat]

Hi there,

My partner, Hanyu Zhang, and I have created a colour transformation visualization system as a school project and we're looking for some feedback from some real world users (ie photographers who do a lot of printing). The idea is pretty straight forward. When you print a photo a colour transformation occurs, based on the embedded icc profile of the image and the output profile of the printer/paper. General convention in colour management is to soft proof your image before you print. This is usually done to make slight adjustments to the photo that optimize for print. However, using soft proofing and/or gamut warning to choose what tools (paper and printer) to get the best print is difficult. Differences in soft proofing (between two different profiles) are subtle, intensive investigation is required to figure out what profile (ie printer/paper) yields better results.

What our tool does is helps to visualize the change that occurs when an image is transformed into a different colour encoding (that is, with the application of different icc profiles), which would aid in making choices about printer and paper.


Also, one thing I should explain is that the visualizations show the magnitude of the distance moved of each pixel in LAB space. So the dark/black areas show that that colour has moved very little in LAB space through transition. The brighter, more yellow, the pixel in the visualization, the greater the movement that has occurred in LAB space. Since CIELAB is perceptually uniform, these movements give a somewhat accurate account of the change in the photo. Its important to note that this is regardless of whether the application is colour managed (obvious soft proofing scenario would differ) or whether your monitor is calibrated or not. Think of gamut warning in photoshop. It tells you where in the photo the colours are out of gamut for a particular profile, but does not tell you how far out of gamut. When operating in Absolute Colorimetric mode, our application is essentially acting like gamut warning, but showing you how far out of gamut one area is compared to another.

The Luminance visualization is slight different, in that it shows only movements in the L* channel, but positive and negative (shown by blue and red scale).

To demonstrate the abilities of the application we've create 3 demo videos which I will explain below. All these tests are explicitly based on the icc profiles (rather than actual prints).  Give to give a little word of warning, the demos do not contain audio, so make sure you view them larger than the default youtube video size so you can see what is going on.  

Case 1: Choosing the "best" paper for the job. ("best" is associated with least amount of change in colour overall, we realize that there are other merits to paper quality that we do not take into account).
Printer: Epson 2400
Papers:
Hahnemuehle Matte FineArt Smooth Photo Rag Bright White
Hahnemuehle Matte FineArt Smooth Photo Rag Ultra Smooth
Epson Ultra Premium Presentation Paper

http://www.youtube.com/watch?v=Mz3j-lNUEi0

The video is unscripted, so here is a basic idea of what's going on.

1.User uploads paper/printer profiles
2.User uploads a photos
-Visualizations are created
3.User filters on the colour green (you would do this if your main concern was colour shifts in green)
4.Mark Photos that are in the top 25% (least change) with a star.
5.User compares the visualizations at 100 percent
6.User compares soft proofs.

Case 2: Which printer performs "better" on a specific image.
Printers:
Epson 4800
Canon 9100 (note that Hahnemuehle profile for Baryta is the same for the Canon 5100, which is in the same price range of the 4800)
Paper: Hahnemuehle Glossy Baryta

http://www.youtube.com/watch?v=uBmjpNDAZDg

1.User uploads paper/printer profiles
2.User uploads a photos
-Visualizations are created
3. User changes rendering intent to relative colorimetric from perceptual
4.User filters on the colour magneta
5.Mark Photos that are in the top 25% (least change) with a star.
6.User ranks photos by least change.
7.User compares soft proofs at 100 percent.

Case 3: L* channel movement of LAB (contrast evaluation for black and white?)
Printer: Epson 4800
Paper: H Baryta

http://www.youtube.com/watch?v=3d63XwRJZBQ

1.User uploads profile
2.User uploads photo
-Standard visualization is created
3.User changes filter to L*
-New visualization is created. The LAB channel filters are special in that they indicate the direction of the movement (see the color scale, although I think this might be backwards).

The main test photo can be found here: http://webhome.csc.uvic.ca/~cgat/ColouredWindowAdobeRGB.tif (the saturated version of the original is the left window).

Before you start commenting on how photoshop shows a different gamut warning as compared to the video, they seem to be pretty consistent when we threshold the amount of change by a certain amount. This is especially so for absolute colorimetric. The other rendering intents different slightly, but not majorly (due to the different colour management engines perhaps?).

Feedback Questions:

Is the premise to the argument valid?

Would you find a tool such as this useful?

What did you not understand in the interface?

How could it be improved?

What do you think of the colour scale (both standard and LAB type)?

Can you see any other values applications of such a program?

Any other general thoughts:



Your feedback is really appreciated. I'll try to respond, but school will be crazy until Wednesday (my last deadline), so you might have to be patient.

Thanks in advance,

Chris Gat (chris.gat@gmail.com) and Hanyu Zhang

PS I should mentioned that negative feedback is welcome as well as positive.

[neile]

Thanks for posting such a fascinating project for us to comment on. Very interesting work!

Is the premise to the argument valid?

I'm not clear on what specifically your argument is, but I'm not sure it's useful to see how far out of gamut a colour is. If it's out of gamut it's out of gamut. A little or a lot doesn't matter (does it?), because it won't print as you expect. I suppose it could tell you which could be brought back in by a slight adjustment?

Would you find a tool such as this useful?

I find the side-by-side comparison very useful. I'd love to see a side-by-side comparison not only of different profiles and different printers but also of different rendering intents (perceptual vs. relative).

What did you not understand in the interface?

Nothing really, it seemed pretty straightforward from the video.

How could it be improved?

Show rendering intents side by side.

What do you think of the colour scale (both standard and LAB type)?

I'm not sure what you mean by this question.

Can you see any other values applications of such a program?

Not off the top of my head.

Any other general thoughts:

Thanks for sharing this with us!

[chrisgat]

Hey Neil,

Thank you very much for the response.

To answer your first question, try to look at it this way. A pixel contains a colour.  This colour is within the gamut of the embedded profile that your photo is using (AdobeRGB for example). When you print this pixel, you ideally want it to stay the same colour, to correspond with the adjustments that you've made in, say, photoshop. When you print, a colour that is considered out of gamut will be mapped automatically (via the color management engine and the icc profile provided) into the gamut of the printer profile.  So if your pixel is extremely out of gamut, there will be an extreme shift in the final perceived colour.  If the pixel is out of gamut, but only very slightly, the perceived shift in colour is actually very small.  This is in the case where the rendering intent is absolute colorimetric (other rendering intents affect the in gamut colours as well). Let me know if you need more clarification.

Thanks for posting such a fascinating project for us to comment on. Very interesting work!

Is the premise to the argument valid?

I'm not clear on what specifically your argument is, but I'm not sure it's useful to see how far out of gamut a colour is. If it's out of gamut it's out of gamut. A little or a lot doesn't matter (does it?), because it won't print as you expect. I suppose it could tell you which could be brought back in by a slight adjustment?



Would you find a tool such as this useful?

I find the side-by-side comparison very useful. I'd love to see a side-by-side comparison not only of different profiles and different printers but also of different rendering intents (perceptual vs. relative).

What did you not understand in the interface?

Nothing really, it seemed pretty straightforward from the video.

How could it be improved?

Show rendering intents side by side.

What do you think of the colour scale (both standard and LAB type)?

I'm not sure what you mean by this question.

Can you see any other values applications of such a program?

Not off the top of my head.

Any other general thoughts:

Thanks for sharing this with us!

[Ernst Dinkla]

Hey Neil,

Thank you very much for the response.

To answer your first question, try to look at it this way. A pixel contains a colour.  This colour is within the gamut of the embedded profile that your photo is using (AdobeRGB for example). When you print this pixel, you ideally want it to stay the same colour, to correspond with the adjustments that you've made in, say, photoshop. When you print, a colour that is considered out of gamut will be mapped automatically (via the color management engine and the icc profile provided) into the gamut of the printer profile.  So if your pixel is extremely out of gamut, there will be an extreme shift in the final perceived colour.  If the pixel is out of gamut, but only very slightly, the perceived shift in colour is actually very small.  This is in the case where the rendering intent is absolute colorimetric (other rendering intents affect the in gamut colours as well). Let me know if you need more clarification.

Nice application.

Soft proofing could certainly be improved. There are however some (arbitrary) differences between rendering solutions in applications. A stand alone program to soft proof images on that later will be printed through another application is tricky. Then there are the differences between the soft proof side of a profile and its printer side.

The extra option of comparing different renderings as suggested by Neile could be another way to solve out of gamut color issues. From Colormetric along Realtive Colormetric to Perceptual the compression of out of gamut colors increases. It could tell whether a rendering choice could be the good compromise or shifting some colors in Photoshop could do the trick. The out of gamut mask (heatmap) doesn't tell much like it doesn't tell much in other applications. The ranking is new and is a nice feature but a more informative visual feedback on what is shifted may enhance it.

The application as shown so far is a nice tool for people with more printers and more papers but I do not think that is the majority of printer owners. So adding a feature like suggested by Neile could increase the value of the software for a wider group.


met vriendelijke groeten, Ernst Dinkla

Try: http://groups.yahoo.com/group/Wide_Inkjet_Printers/

[marcsitkin]

I did a quick read on your project, and think it has merit. I wonder if you are familiar with Colorthink from Chromix, which I believe may cover some of the same ground?