Printer Profiling on the Cheap for the Amateur Photographer

Version 1.0 (June 2002), ©2002 by Dale Cotton, all rights reserved

Intro

Over the past decade the advent of personal computers with graphical interfaces, Photoshop, film scanners (and/or digital cameras), and Epson’s stunning inkjet printers has wrought a revolution in the output half of colour photography. Previously colour photographers were much better equipped to take pictures than to turn those pictures into fine art prints.

But those of us who have invested in the digital light room (as my friend Andy calls it) soon begin to realise that there is yet another component necessary to achieve desirable results, and that component bodes to be pricier and more complex than everything else put together. I refer to calibration and profiling.

There exists a class of equipment involving optical spiders and colour targets and flatbed scanners aimed at resolving this issue. Unfortunately, the consensus is growing that the affordable options are less than adequate and the adequate options are less than affordable. Another solution is to purchase custom-made profiles from places like Chromix, which may be just the ticket for many.

This tutorial is intended for a very narrow audience. For it to be useful to you you would need to meet these criteria, you must:

1. Live in poverty (possibly from spending too much on photo equipment)
2. Print to an Epson inkjet printer (but may well work with other brands)
3. Require an extremely accurate colour match between printer and monitor but understand that paper and phosphors are ultimately irreconcilable
4. Own and use Photoshop with some aplomb
5. Possess a reasonably good eye for colour (or have enlisted the aid of someone who has a good eye (perhaps a local art student even more impecunious then yourself)
6. Possess the patience of a saint (or the desperation of the damned).

Is there anybody left reading this? Being optimistic by nature, I’ll presume there is at least one reader out there and move on.

Figure 1. A Curves dialogue profile (Epson HW Matte).

Note: Feel free to try my own curve, shown above. Click here to download it. I created this for Epson HW Matte but it's darn' close for the other Epson papers I have on hand. If you have a 1270 or 1280 et me know what results you get - I'm curious how close one printer is to the next.

The method I used to profile my printer involves no cash outlay beyond maybe $20 in paper and ink. Normally, a device profile is a special type of text file with an .icm or .icc extension. The above-mentioned expensive equipment and software is needed to generate these profile files. What I do is to create a Photoshop curve file (.acv) using Photoshop itself, then apply that curve to each image just before printing. Details follow.

Before you get to that stage there are some preliminaries:

1. You need to have your monitor calibrated and Photoshop properly configured. The affordable way to do these things is to read one of Ian Lyon’s superb tutorials, called Photoshop 6 & Colour Management, on his Computer Darkroom site. When reading that tutorial, calibrate your monitor according to Ian’s instructions using Adobe Gamma. Here’s a hint: have open and visible a grey scale image, such as the top portion of Figure 2, when you’re adjusting the red, green, and blue gamma squares. Aim to get the greys as pure as possible – not reddish grey or bluish grey or yellowish grey, but just pure grey grey. You may want to deliberately drag the red, green, and blue squares off-centre to see what the various shades of impure grey look like.
2. Next, follow Ian’s Example 1 method for configuring Photoshop for printing. If you normally, use some other editing space than Adobe RGB, set that as your working space.
3. Next, you need several sheets of the same paper you wish to profile. For example, I normally print with Epson HW Matte in A3 or Super A3 size, but for testing I used Epson HW Matte in 8.5 x 11 size. (And, of course, you’ll need ink!)
4. Next, you’ll need to work in the same lighting you wish to view your prints in. In fact, what you are really doing here is profiling a combination of paper, ink, and viewing light. If you want to do a run of prints for viewing in tungsten light, then create a tungsten light profile. If you want to do a run of prints for viewing in indirect sunlight, create an indirect sunlight profile too.
5. Finally, you’ll need to save Figure 2 to your hard drive and open in Photoshop. (Convert Figure 2 to your preferred working colour space if necessary.)

And some caveats:

1. Using this method of profiling there are two colours, pure magenta and pure royal blue, and shades extremely close to them that I simply cannot correct (I only print to Epson paper). I’m hoping someone more astute then myself will read this, laugh, then write to tell me what I’m doing wrong. So far I’ve only come across one image (Figure 11, below) that has either of these colours in it (pure magenta).
2. Since colour space is three dimensional and Photoshop curves are two dimensional, you may come across two different hues that are almost exactly the same brightness yet which require two different red settings, or two different green settings, or two different blue settings. Here your job will to be to find a compromise you can live with.
3. This method of profiling does not have the conceptual elegance of the operating system level ICC system. In practice this means remembering to add a printing curve before printing rather than choosing an ICC profile before printing.

You have been warned. Since I’m giving this methodology away for free, I have no hidden agenda, so I’ve laid the bald eagle truth on the table. <g>

Figure 2. Dale’s Super-Duper Colour Test Target (Now! Newly revised and expanded! ).

Now you’re ready to profile.

Profiling with Curves

• Save Dale’s Super-Duper Colour Test Target by right-clicking on it or the Mac equivalent, then open it in Photoshop. (If you have the Macbeth Color Checker image or similar, feel free to use it instead or as well.) Print it using the following routine:

Figure 3. Print Options dialogue.

• Load a sheet of the paper you’re testing with in your printer.
• Press Ctrl+Alt+P (or Command+Option+P?) to bring up the Print Options dialogue (Figure 3).
• Click Page Setup... if the paper is not the correct size or the orientation is not landscape.
• Return to Print Options, Change Position to the upper left corner. (Leave scale, etc. alone.)

Figure 4. Print dialogue settings.

• Click on Print..., make sure Source Space is Document and Print Space Profile is Printer Color Management.

Note: If you have a previous ICC/ICM profile from whatever source for the paper and ink combination you are using, it would be interesting to use it "beneath" the curves profile you are creating here. In theory, it should make your job that much easier. In that case, use the Print dialogue settings you would normally use with that profile instead of the ones I am suggesting.

Figure 5. Advanced dialogue settings.

• Click on Setup..., then Properties..., then Advanced. Set your paper to the correct choice (in my case HW Matte), set Print Quality to 720 to save on ink, and set Color Management to ICM. (You may want to Save... these settings for future use in profiling.)
• OK back to the Printer dialogue and click OK to print.
• When the print is done hold it up to the monitor and compare each colour patch between the two. This is your starting point. Many should be exact matches. If all are exact matches, you’re done - but this seems unlikely.

Figure 6. Layers palette.

• Now create a Curves adjustment layer named Printer Correction. Create a second Curves adjustment layer named Trial (see Figure 6).
• Choose any patch that does not match between print and on-screen image (I suggest one of the grey patches) then switch to the Trial Curves layer. Move the Curves dialogue so your chosen patch is not covered on-screen.

Figure 7. Clicking on a colour patch from within Curves.

• Click on the offending patch and hold down the mouse button. You’ll see a little circle with cross-hairs on the patch and another circle on the Curves diagonal. This marks the brightness of the patch. If the print patch is darker than the on-screen patch grab the RGB channel curve at the point that was circled, then drag it up and left until the print and on-screen patch are the same brightness. If the print patch is lighter drag down and right.
• If the print patch is still a different colour than the on-screen patch try to guess whether it has too much blue and/or too much green and/or too much red. Let’s say you guess too much red. Switch to the Red channel curve (the channel selector is circled in Figure 7), then drag it down and right (from the same point on the diagonal you used in the RGB channel) to varying degrees until the print patch and on-screen patch look as close as can be got.

Note: see Appendix 3 to practice this technique and develop your colour-correcting eye without wasting paper and ink in the process.

• Repeat for the other two colour channels (green and blue if you’ve already done red). Keep playing with this until print and screen look identical.
• Now you need to invert and transfer these changes to the Printer Correction curve. Back in the Trials curve, notice the two numbers – Input and Output. Write down those numbers for each channel you’ve modified. For example, if you changed the RGB channel to input 187 and output 196, write down RGB: 187,196. If you also changed the red channel to 111 and 103, write down R: 111,103. If you also changed the blue channel to 151 and 157, write down B: 151,157.

Figure 8. Inverting a curve by reversing Input and Output.

• Now switch to the Printer Correction curve. Given our example in the last step, type 196 in the Input and 187 in the Output – the opposite of whatever you wrote down from the Trial dialogue. Ditto, you’d switch to the red channel and type in 103 to the input and 111 to the output. Do the same for each channel you modified.

Note: You will only be able to place a limited number of correction points on a curve by the numbers. A subsequent problem patch may require a change to the same (or nearly the same) point on the curve as an existing correction point. In that case I just eyeball the displacement between the correction point on the Trial curve and the centre line, then add that much displacement in the opposite direction to the displacement already present on the Printer Correction curve. If you're off by a tad you'll catch it in the next round of adjustments.

• Repeat the above five steps for four or five other offending patches. When you place a second correction point on a single curve line, notice that you’ll have to first click on the line anywhere but where the original point was before typing in the new numbers.
• If any channel curve has a few (but more than one) points on it, you may have to create more correction points to drag the portions of the curve that have no points back to the original diagonal position. Photoshop believes in the "every action has an equal and opposite reaction" philosophy.
• In the Printer Correction curve dialogue save the curve as Printer Correction 1. In the Trial curve dialogue, click Cancel while holding down the Alt/Options key to reset it to neutral.
• Finally, turn off the Trial curve’s visibility and turn on the Printer Correction curve’s visibility (by clicking on the appropriate eye icons circled in green in Figure 6).
• Now repeat from the top to re-load your sample print in the printer. Use Print Options to locate the next image of the Colour Test Target in a blank place on the page. Then reprint it. After then modify offending patches, then save your Printer Correction curve as Printer Correction 2, etc. I find it helpful to label each new image on my test print page as Printer Correction 2 or whatever was used to print it.

Note: it is important to save each new Printer Correction curve you create with a new name and very useful to label each corresponding test print with that name. It may come to pass that your latest trial actually takes you astray from your chosen path. In this case, simply revert to the best previous Printer Correction curve and try anew.

After several iterations of the above you should be getting pretty darned excited about the level of accuracy you’re beginning to see (perhaps for the first time) in the match between print and screen. At some point you need to move from the artificial security of your Test Target insularity out to the "real" world.

Tip: after a few iterations you will likely find yourself zeroing in on just a portion of the Colour Test Target (or other image) at a given time. Save paper, ink, and time by working with a crop of the full image. In fact, I found myself concentrating on just the grey scale for the first portion of the adjustment process.

I took several of my most frustrating images, reduced them to very small (three or four inch) dimensions, assigned my best Printer Correction curve to them, then printed. If I saw any colour discrepancies, I used the image as my test target and followed the same procedure as above. Here are my choicest challenges, but I encourage you to create and use your own:

Figures 9, 10, 11. A few images with which to challenge your best profiles (click for larger version).

Note: As mentioned, I cannot get the blossoms on the crab tree in Fig 11 to print as magenta: I get hot pink every time!

Appendix 1: About the Target

During the formative stages of creating this profiling process I used a copy of the Macbeth Color Target I downloaded from somewhere then modified as my ideas developed. When I decided to write this tutorial, I felt uncertain as to whether Gretag-Macbeth really wanted me to distribute their famous image file. This led me to dream up the idea of making my own from first principles.

Figure 2. Dale’s Super-Duper Colour Test Target.

The first row (shades of grey) and rows 5 through 10 (pure colours) were created by the numbers. If you move your pointer over them in Photoshop with the Info palette open you’ll quickly see the regular (hexadecimal) RGB patterns involved. Row 4 is my attempt at a generic flesh tone in various brightnesses. Rows 2 and 3 are a random sampling of tertiary colours taken from my photos that hopefully provide a good cross-section of real-life hues (Psst! And two or three are even "borrowed" from the venerable Macbeth chart itself).

The first row contains double-sized patches for three reasons. One, many photographers have considerable experience with grey scale tints. Two, most of the rest of us can nevertheless easily see subtle grey scale tints (perhaps because our night vision is in grey scale?). Three, psychologists have demonstrated that the human memory system responds better to colour than black-white-grey.

All rows except 2 and 3 follow a pattern. Each start with a single colour, the one in column D. That colour is made successively lighter (equivalent to adding white) for columns C, B, and A; and successively darker (equivalent to adding black) for columns E - I. For example, D5 is pure magenta; A5, B5, and C5 are lighter/pastel magenta; and E5 and F5 are darker magenta. However, because the starting point (the cell in column D) is not the same brightness, all the cells in any given column are not the same brightness. For example, B6, B7, and B9 are nowhere near equally bright.

The pure colour cells – A1, F1, D5, D6, D7, D8, D9, and D10 – are not under your control. A1 will be the whiteness of the paper you are using. D5, D7, D9, and F1, are the pure CMYK inks. While D6, D8, and D10 (RGB) seem to be heavily controlled by the printer driver, which believes (with a calm certainty that the born again can only aspire to) in the correctness of its algorithms.

Feel free to enlarge, reduce, or rotate this image to achieve the dimensions that work best for you. Also feel free to distribute the unaltered image, so long as you do not charge money for it.

Appendix 2: Printing behind the Scenes

Your monitor is an RGB device that creates all other hues by mixing red, green, and blue in varying strengths (black being zero amounts of all three, white being maximal amounts of all three). This is also known as additive colour. Stage lighting, for instance, is done by mixing red, green, and blue filtered spotlights. Your printer is (likely) a CMYK device. It mixes cyan, magenta, yellow, and black inks on white paper in varying amounts to achieve all other hues. This is known as subtractive colour and is what paint-and-brush artists and school children with crayons or finger paints do.

Your monitor is also a luminous device – it paints with light, which is what additive colour is all about. While the prints that come out of your printer are illuminated, which is what subtractive colour is all about. A brilliant luminous red, for example, is a fundamentally different thing to perceive than any illuminated red can be.

Your printer driver receives RGB numbers as input and translates them to the corresponding CMYK numbers for printing. I presume the translation is based on empirical testing. Two other variables are the whiteness of the paper and the chemical properties of its surface coating, which effects how the CMYK inks mix.

Your printer driver is juggling RGB to CMYK translation, additive vs. Subtractive colour, paper whiteness and coating chemistry to try to create a seamless and ideal printing experience.

For more about colour theory see Light, Color & Human Vision by Miles Hecker.

Appendix 3: Of Casts and Curves

Figure A3-1a & 1b. Sample colour patches.

Save both patches to your hard drive and open both in Photoshop. Click on A3-1a then create a Curves adjustment layer for it:

Figure A3-2. Patches and Curves dialogue at start.

Our challenge is to change the colour of A3-1a to match the colour of A3-1b – henceforth A and B. Let’s tackle this empirically (also known as trial-and-error). Grab the RGB channel diagonal near the centre and swing it wildly up and down. None of these variations do much to make it look like B. Switch to the red channel and repeat. Again, little improvement anywhere. Switch to the green channel. Here we see that dragging the curve steeply down and right quickly brings A into line with B. When you’ve got A as close as you can get it to B using the green curve. Try very small adjustments with the other channels to see if you can get it still closer.

Figure A3-3. Green curve adjustment.

Now, let’s apply a little logic. The mysterious numbers on the bottom of A and B are the RGB numbers I saw in the Photoshop Info palette when I created each (the JPEG compression process has since played a bit of havoc with their purity, but that’s real life). We can see that the G, or green number in each pair is the one with the largest difference, and that’s just what we found empirically – a drastic change to the green channel curve was needed to match the two colours.

During the profiling process we don’t have the advantage of two sets of RGB numbers. We can use the Info palette to see the RGB value of each on-screen colour patch; but we’re missing the corresponding RGB numbers from the print. Instead, we need to train our eye to recognise colour casts. Compared to B, A could be said to have a severe green cast. Conversely, compared to A, B could be said to have a severe red cast.

Figure A3-4a & 4b. Two more sample colour patches.

For practice, repeat the above with Figure A3-2a and A3-2b. Don’t cheat by looking at the RGB values until you’ve tried your best shot using your eyeballs. Cheaters never prosper.

You can systematically train your eye to see colour casts. Simply open Dale’s Colour Test Target, add a Curves adjustment layer, then drag each of the channel curves up and down while looking at a random colour patch to see the effect of adding and subtracting each primary colour is. This skill is not just useful for profiling on the cheap. Nearly every photograph straight from the digital camera or scanner will have a colour cast. This can come from the hardware, the film or digital sensor, or the lighting. Sunlight, for example, is reddish in twilight, greyish or bluish when overcast, and bluish when the sun is overhead. A typical outdoor photo has some degree of blue cast; removing that cast will typically create a much more pleasing image. Or say you’ve taken a portrait of someone in a bright green shirt. Amazingly, light bouncing off the shirt and onto the subjects skin can create a subtle green cast on the affected parts of the subject’s skin. Again, removing that cast can keep the subject from giving a subliminal impression of seasickness!

Black-and-white photography is dramatically simpler than colour photography simply because the shades of only two colours (black and white) are involved – a very limited palette. Colour photography is more complex but richer because of its broader palette. Just as a musician must eventually go beyond simply learning notes to controlling the tone/timbre of his instrument, so a colour photographer needs to master the rich world of colour.