Wednesday, 25 January 2017

Paper to Pixel to Paper Again: Part 3

1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 9 | 10 | 11 | 12 | 13 | 14 | 15
16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | 25 | 26 | 27 | 28 | 29
A guide to creating the best looking line art in print in the new digital print world
Part 3
Sharpness & The Selection Of A Sweet Scanner


This is the third installment of Paper to Pixel to Paper Again, a series that explains (in an overly thorough manner) the how-to's of preparing line art (and later in the series, color art!) for print.

If you haven't done so yet, please go back and read the first and second installment.

Really. Please do. 


When we left off last week, we were discussing how you NEVER EVER NEED TO HALF-TONE LINE ART, and how not half-toning allows you to work in a much finer resolution space. 

As I mentioned last week, all of the Cerebus Restoration projects have been delivered to the printers as 1-bit (line art) 2400 ppi originals. (In a later installment I'll get into why I arrived at that resolution, and all of the benefits we'd had as a result. But for now, let's just take this fact for granted.)

But just because we're delivering 2400 ppi bitmaps doesn't mean we're scanning at that resolution. This wouldn't be practical or even desirable, for a few different reasons.

Instead we've been scanning at different resolutions (and different formats) depending on the source material, and then upscaling and finally sharpening to arrive at the delivered size. I'll spell all of the how-to of this out in the next installment. 

For the most part, there are three sources for materials in the project so far—print materials, photo negatives used to produce the printing plates in the initial run of the book, and original artwork. The print materials and the photo negatives are both at size (or close to the size) of the printed books they will be used to generate, so after much testing, these have been scanned at 1200 pixels per inch. The original art, however, is much larger, and so for the most part these can safely be scanned at 600 pixels per inch and retain all of their detail in the final result.

(How much larger? For most of the run of the book, the active image area of any given printed Cerebus page was around 6" x 9", whereas the corresponding area of the original art was around 10" x 15". For most books the original art was reduced to 60 percent of the original size, with the exception of the latter half of Church & State I through Jaka's Story, which was reduced to 56 percent instead, presumably to try to address any future problems with binding etc in collected editions of the books.)

And why try to skimp on resolution at the scanning stage? There are a few answers to this, but the biggest issue is sheer time. Every time you double your scanned resolution, you're quadrupling the actual time it takes for the scanner to scan the image. (It's not precisely quadruple, as there's a certain amount of setup, warm-up, etc that remains the same no matter how fast the scanner operates).

We're already dealing with a six thousand page problem. Quadrupling your scan time for no good reason isn't ideal.

As to why, exactly, you can safely upscale so much for line art applications when it remains such a general no-no for color reproduction — the very short version is, the color scan contains much more information per pixel than the resulting line art pixel, and that information enables the upscale and attendant bitmap conversion to be very precise to the source document despite the resolution change. More (I seem to be saying this a lot?) in a future installment.


Okay, you've drawn your masterpiece, you've read way too much information about resolution and line art, and now you're impatient to get on with it. Now how do you get all your inky dinky little lines into this here computer box?

Shooting line art used to be a complex process, involving a huge stat camera loaded with orthochromatic film, and an expert operator with a good eye and a ton of experience. 

No, uh, not quite that large.

Now, the same goals can be accomplished with a good flatbed scanner and... well, still a lot of skills involved. Hopefully most which can be acquired by reading this series...

Because it was used to both enlarge and shrink artwork, and because every other stage of printing would contribute to the inevitable fill-in of line work, a good stat camera was prized for the resolving power of its lenses. The sharper an optical system is, the more detail is captured from the source, and the less that capture needs to be manipulated to retain that detail. This is still true when we're talking about digital sources.

With that in mind, what follows is a run-down of many of the commercially-available flatbed scanners out there, and what the problems and advantages of each are.

Picking Your Scanner Box Thingie

The Cerebus Original Art Dragnet project has really given me some interesting insight into the wide variety of scanner types and qualities out there. Although the art changed drastically over the years of the book, many of the elements and particular areas of concern from a production standpoint remained the same, meaning that I've had the opportunity to see the same types of screen tones scanned on a huge variety of devices. 

If you really, really want to know the optics of flatbed scanners, what's actually happening under the hood and how to really measure the bogus stats that are spewed out by the manufacturers of these products, then visit the fantastic ScanDig and read some of their film scanner reviews. I am indebted to their site and the information they so casually drop in their reviews. And yes, they're selling their own services and there's undoubtedly some self-interest involved in their critical scanner reviews, but they have an incredible bank of knowledge, a professional's eye and a true enthusiast's passion. Good reading. 

The Bad News First

There are very few large-format flatbed scanners available, and none that I know of in a middle-range price. And the most affordable ones are, ahem, total crap.

Yes, this means I might insult your scanner. Please prepare yourself.

Mustek ScanExpress A3 2400, or 1200, or any other Mustek

It's hard for me to imagine how a company like Mustek continues to exist in an age where people can read and publish public reviews. They make a demonstrably inferior product that doesn't function as promised, with incredibly buggy drivers, and offer no support or returns. How, you might ask, do they stay in business? Because the prospect of paying $339 for a "large format" flatbed is too tempting for poor cartoonists who just want to scan their damn pages.

Don't do it. Please. Don't do it.

The Mustek uses a special (i.e. super cheap!) kind of scanning sensor that's only capable of seeing with an extremely narrow depth of field. How narrow? If you're scanning a watercolor painting, any slight buckling or indentations in the paper from moisture will be OUT OF FOCUS. Literally the texture of the paper will be out of focus. If you don't load down the top of the scanner with books, the edges or any microscopic lifting at all will result in completely out of focus scans.

Making things worse, the scanner plate is not actually the advertised 11" x 17". Making things even worse, it features a special raised lip that causes any artwork larger than this "almost A3" size to lift off the surface, and thus, to be, you guessed it, completely out of focus.

Add to this motor jitter that causes random lines of the scan to be misaligned with the previous ones, and you have a completely useless paperweight.

UNLESS you only scan almost-11 x 17 artwork that's always completely flat, that has no mechanical tones, you keep some bricks around in your studio to hold the lid down, and you can stomach the idea that this thing will die every few months and need to be replaced, if not returned immediately because they don't power up. Then, hey, do ahead!

From this view, the pages is looking good! Alright, got a good scan. Let's check closer in.

Looking pretty good! Pretty sharp, with no visible artifacts that would indicate any unwanted software sharpening being applied. Maybe this thing really will work out.

Oh, hey, what's happening at the top of this scan??


As you might imagine, having a scan that is overall a little blurry might be okay, given that this is going to end up as line art. Having a scan that is PROGRESSIVELY BLURRIER is next to useless.

.... and that's my time for this week! Next week: scanners continued. Dah dah daaaaaaaahhhhhhh!

Sean Michael Robinson is a writer, artist, and musician. See more at


Sean R said...

Hey Dave! In response to your comment from last week: thanks for noticing :) I will discuss some publishers that are definitely doing things well (at least with certain books) in later installments...

Dave Kopperman said...

About ten years ago, I bought myself the Microtek 9800XL - about $1500, as I recall. Flatbed around 12x18. Only problem is that the company almost immediately withdrew support for it in the US for several years, so any OS update is a real nail-biter. Not bricked yet!

Unknown said...

Dave - Thanks for posting! Anyone who has a large flatbed scanner out there, don't be shy.

Sean - That was going to be my next question: are there any FLAWLESS publishers? And if there aren't, what is up with that (would you guess)?

Carson Grubaugh said...

I have an Epson Workforce-7610 printer-scanner combo. The scans aren't the best, when you really zoom in you can see where the image is being stitched together, or something, but the whole unit was $150 bucks and it works very well for the trace, print, draw, scan, print, process I am using.

Travis Pelkie said...

That's a very BWS head, isn't it? Funny what you notice when everything is so close up and good looking!

Unknown said...

Travis - Yes, BWS inked by Bill Sienkiewicz which is something to conjure with.

Dave Kopperman said...

Maybe it's a function of my age, but I've always found Sienkiewicz's inks (circa '84 on, when he really loosened up) to be much more visually appealing than BWS from almost any era. I feel like Smith's pencils have so much lightness and energy, but then his inks are static and heavy.

Damian T. Lloyd, Esq. said...

This stuff is gold. A collection of these articles and Dave's own Self-Publishing Guide -- plus something about WWW publishing -- and you've got a complete manual for the aspiring cartoonist.

-- Damian

Dave Kopperman said...

Damian - you should check out "Drawing Words and Writing Pictures" by Jessica Abel and Matt Madden. Definitely the most comprehensive soup-to-nuts guide for the aspiring to publish cartoonist, including really good chapters on scanning for line art (which align with Sean's ideas) and self-publishing. Also the best description of how to use an Ames Lettering Guide I've seen!

Sean R said...

Hey Dave,

I think a response to that will get its own installment!


Thanks for the kind words. I'd appreciate any sharing/link spreading you can do to people or venues who might be interested.

Dave: I read that when it came out. Found some of the after-chapter exercises went over really well as classroom assignments (presumably what the book was intended for). I don't remember the tech chapter having anything about sharpening though, what I'd consider the secret sauce! I'd be curious to go back and look at it after finishing this series...

Dave Kopperman said...


It's definitely nowhere near as thorough as what you're putting together - which is looking like it'll be the definitive text on the subject - but I remember it as being the first place I saw recommendations for using the Threshold tool (I've been a Photoshop professional for two decades and have literally never had call to use it!) and saving as .bmp. A quick look shows some areas of disagreement - 600 dpi grayscale scans, later converting to 1200 dpi - but you can tell their thought process is similar enough to yours that it stands out from many other sources I've seen.

Annukka said...

About the techniques in Jessica Abel's book: I once wrote to her about the up-res of 600 ppi scans to 1200 as I found it strange advice at that time :) She kindly answered: "this is upresing, but the difference is that grayscale is 256 colors, and bitmap is only 2. So if you're upresing grayscale, before thresholding and going black and white, you've got plenty of data, way more than enough, to make a nice smooth line.
Does that make sense? Do a few tests, you'll see. scan 600 bitmap, then 1200 bitmap. then 600 grayscale, then upres and threshold. You'll see that the two 1200 files are virtually identical, but you've got the advantage of a faster scan time for grayscale, plus all the editing tricks you can do in grayscale and not bitmap."