The Forensics of Exposure

SEAN MICHAEL ROBINSON:

Hello all,

It's been an eventful week at the Cerebus Restoration Fortress of Duitude. Mara has shifted from indexing and scanning materials to cleanup and layout, and is doing an outstanding job so far. We're timing ourselves as we go, getting faster even as our attention to detail improves.

But the more you can perceive, and the more correction you're capable of making, the more possibilities present themselves.

A small illustration--

Mara has been asking me from the very beginning if there are existing ways to check a page's exposure with forensic accuracy. That is, given the consistency of Cerebus' 30 percent dot tone, can you actually measure what that tone should look like in print?

This is a much harder question than it might seem. Yes, there are ways to measure ink coverage on a page. But for a variety of reasons it's still somewhat of a crap shoot to measure tone at the output (print) stage, the chief hurdles being dot gain, and in the case of fine tone, optical dot gain (Yule-Nielsen Effect) as well. The shorter version-- printing causes expansion of ink, and that expansion is a non-linear process, and isn't going to be reliably the same from one system to another. In addition to this, your eye perceives different sizes of tone as different optical densities. A coarse 10 percent half-tone will appear lighter to the eye than a very fine ten percent tone, even without actual dot-gain.

TLDR-- it's hard to trust your eyes.

Combine this with wildly varying sources-- newsprint, photo negatives, original art, all in different conditions, scanned under different circumstances-- and you have a Gordian Knot of exposure.

But as my ability to manipulate this exposure in-file has improved, so has my ability to distinguish the exposure problems from each other, always aided by the constants (or relative constants)-- Cerebus's dot tone, Dave's lettering width. 

So what's the anchor, then?

Well, Dave and I could agree on what the best representative sample of tone was from the recent Lebonfon test printing. So could I use a sample of that, flying it around to test exposure on new pages?

Works pretty well. How about making some digitally pure 30 percent tone to test? 

Still works pretty well, but we're still talking about using my eyes for something there should be an objective measure for. Audio engineers know that their ears get fatigued over a day's worth of recording or mixing--your high end perception starts to change over time, and if you let yourself go too long, you'll wake up the next morning to find you've made an ear-bleedingly bright mix. So instead, you take breaks, you listen to reference music to re-calibrate your ears, you take a look at a graphic equalizer and check out the frequency distribution.

That's what I need, I thought last night. The equivalent to a graphic equalizer. Sure, you can use Photoshop to measure gray density, but if only there was a way to do the same for half-toned material.

I woke up this morning and immediately knew the answer.

Here's a Cerebus figure from page 170 of High Society. His tone looks good in print, and is comparable in-file to the page Dave and I agreed is spot-on in the test signatures.

I've applied some gaussian blur to a segment of tone, with a radius greater than the pixel width of the tone itself, causing the areas to average out to gray. Now that Cerebus's gray tone is an actual continuous tone image, I can use the eyedropper tool and the Info window to analyze it-

And here we have our forensically accurate answer. The Cerebus tone for this particular figure is occupying 28 percent of the white space. With just a teeny bit of room for expansion when it hits the page, it's no wonder this tone is looking optically perfect in print.

After discovering this, I did a quick check of the files that created the test signature, and sure enough, the results of this test align perfectly with my suppositions about the material. The figures Dave and I had deemed "just right?" Between 28 and 31 percent. The ones I had deemed "close but too dark?" 34-36 percent. The one page with very damaged tone that was way too dark? 40 percent. A page that I had done at the end of the day last week, and now deemed too light? 26 percent. How about the issue 5 tone from the replacement signatures, the one I thought was spot-on? 30 percent.

I am very pleased to say I now have my objective measure. I just put together a script and short key for it-- it'll take five seconds to test the exposure when I'm working a page.

It's very possible that someone reading this is now nodding their head, saying, "of course that's how you'd do that." If that's the case, I wish you had told me. It's the kind of solution that seems completely obvious in retrospect, but took me weeks to arrive at without any kind of direction. But once again, it was my increasing perceptual accuracy that drove me to search for a more accurate testing method in the first place.

Let me go ahead and state this publicly--when we're done with this restoration, one year, two years from now, I'm going to write up and share every technique that I've learned, try to help other people arrive at solutions for these problems and see if I can't help expand the common knowledge a bit. It's been a little hard for me at times, essentially learning these skills in public, but I hope that the return is worth it.

Back at it next week! Thanks for your time everyone!