Other Scanning Old Magazines - a Worklog

Because the quality is terrible. I mean, it's not, it's *okay*, but it's nowhere near what you get from a scanner.

I actually have a CZUR book scanner: a phone-like camera on an overhead stand with lights and lasers that measure the curvature of the pages to dewarp them. I don't use that much, either, for the same reason, and I can use that hands-free - it has a foot switch.

Plus, the scanner is much faster - I can't snap phone pictures at 80 images a minute - *if* the bloody pages feed straight...

EDIT:
This is what I mean:



Looks OK at first, but then you realise a bit of the top is missing and the dewarp algorithm has got very confused and made the bottom, which was already flat 'cos it's the front cover, go like a flag.



Page is still warped towards the spine, the glossy paper has a reflection from the window, and my thumb's in the way. Bottom-left corner's blurred, too.



Same issues again: the page is too warped for the dewarp algorithm to handle, and my thumb's in the way. It's also cut off the top again.

As for quality, I have to admit it did a better job than I was expecting (and the last time I tried similar software, which was, granted, a few phones ago now) - but it's still not what you get from an actual scanner.

Here's the best shot Genius Scan managed:



Which is, yeah, it's readable, and for anyone lower on the autism spectrum than I it's probably absolutely fine. But when I know I can get this:



Then, yeah, I'd like that, please!

(Both 1:1 crops of the unedited image: Genius Scan was set to "Highest Quality" and the scanner to 300dpi; the scanner goes up to 600dpi, but I'm already running out of RAM on the bigger mags and spitting out 700MB PDFs, so I'm sticking at 300dpi. Helps that it's about three times as fast as 600dpi, too.)

I've picked up a new scanner which, I'm hoping, will more reliably feed the pages straight *and* is faster at 80ipm instead of 50ipm (double-sided); I can't do much about how long it takes to cut the pages free from the spine, though, without risking a finger or two...

 

What stage is my obsession at? It's at the "comparing JPEG encoders" stage, thank you for asking.

 

Finished cleaning the last of the mouldy ones today. It'd be nice to do the other two bags of not-visibly-mouldy-but-a-bit-stinky ones too before they get sliced and scanned, but my back's already killing so that's enough for today.

New scanner hasn't arrived yet, but I gave the pinch roller on the old one a good clean and with the new page separator in there it's behaving itself for now. Even did an entire Software Warehouse catalogue without needing a rescan!

That's Issue 36 to PNG - I wanted to do that first 'cos not only was it one of the mouldy ones but it had water damage to the back corner. Gave it a good douse in isoprop and dried it out before scanning. 756 pages including covers... 13.6GB(!).

Not sure I've got it in me to do another today!

 

Anyone still rocking a Matrox Millennium II in the 21st century?

 

Still no sign of the new scanner. It was booked in at a parcel shop yesterday, but the tracking doesn't say it's been dropped off yet. Bah!

Been busy with work of late, so I've been letting things go off the boil a little. Decided to rectify that: scanned Issue 89. Cut the top edges at an angle again, I still haven't worked out how to stop that. As always, it only affects the edge-to-edge catalogue inserts - speaking of which, thanks, Dabs, for picking some kind of weird glossy paper that's gone really wrinkly and wouldn't feed through the ADF. No casualties this time, and it was relatively short so I did it manually on the flatbed.

380 pages, 6.6GB.

EDIT:
Doing issue... 116 now. We're officially moving out of "nostalgia" and into "hey, I still use that."



I had that on my camera literally this morning, doing product photography!

EDIT EDIT:
Disaster: I'd literally scanned the back cover of 116, when my music stopped; the desktop'd crashed. Rebooted, opened up Simple Scan, held my breath... "There is an autosaved book available. Do you wish to open it?"

Yes.

It *looked* to have everything except the last page, but it's not happy: crashed when I scrolled to the end. Reopened it, reopened the scan, and hit export: it got to page 304 (P290 in official numbering), skipped 305, then wrote a zero-byte 306 before crashing.

At least I don't have to scan the whole thing again, but I'm really curious as to what happened there. It's never done that before!

 

I've got (another) new scanner on the way, after the first one arrived smashed: a Fujitsu Fi-7160, which lacks a flatbed but has a supposedly top-end feed system, L-shaped paper path, and intelligent anti-skew - and runs at 120 images a minute, allegedly(!)

There are three parts to the process, though, and they all take a long time: cutting the mag; scanning the pages; tweaking the scans. I'm hoping the new scanner will speed up both the scanning stage (by requiring fewer rescans when something feeds at a bad angle and by scanning the pages more rapidly) and the tweaking stage (by deskewing the pages itself so I don't have to.)

Which just leaves the cutting stage...



Cheap Chinese guillotine, eighty quid. Claims it'll do a 300-sheet stack of A4 pages, or up to 4cm in thickness - and while some of these mags go over 300 pages, they're wafer-thin paper so 4cm is easily twice the size of the average issue.

If this works, I'll be able to slice the spine off in one go instead of standing there for an hour with a steel rule and a Stanley knife. Better still, it'll be straight - no more cutting more off the back pages than the front pages, no more funny angles, just a neat, straight, even cut all the way through.

I'm suspicious of its claims, of course, and I was tempted by one on Amazon that claimed 400 sheets - but looked suspiciously like ones AliExpress had claiming only 200. This one's from Ali, but from a UK warehouse - should be here in three days. Between that and the new scanner (fingers crossed this one arrives in one piece) I'm hoping I can dramatically reduce the time it takes to digitise each mag.

God knows where I'll put the thing when I'm not using it, it's massive.

 

No guillotine yet, but new scanner get!



Threw a hand-cut Issue 90 at it as a test. First impressions: it's too dark out-the-box, and even though I've tweaked it now it could still do with more careful adjustment before I scan more issues.

It's fast. Super fast. Very happy with that. Haven't even installed any drivers, it's picked up by SANE nae bother.

It does not auto-deskew. As expected, I'd imagine that's a feature of the Windows-only bundled software.

It does, however, feed very straight by itself. Way more so than my old scanner. So, way fewer rescans - I had to rescan maybe five sheets this time around, and that was only 'cos I was being fussy - they could all have been fixed in post.

The not-so-goods: turns out Simple Scan will only let you choose which of multiple connected scanners you want to use at the start of scanning; you can't switch halfway through, so I can't scan most stuff with the sheet-feeder and some stuff with the flatbed. Or I can, but I have to save and restart every time I switch scanners. That's annoying, but ameliorated by the fact that the new scanner will happily feed cardstock the old one would fail on - so I can scan the covers and fold-out inserts through the feeder.

Quality seems a bit lower, maybe? That might be 'cos I'm still adjusting the thing, though.

The auto page size detection doesn't work. Like, it doesn't always work on my old scanner, if a page starts with a plain black border at the top it'd often cut it off - but this one is failing to detect the bottom of any page. "Auto" actually seems to mean "a pre-set size slightly smaller than A4." I can work around it by setting it to A4 manually, but it's still an annoyance.

Am I pleased, though? Oh, yes, This is going to speed things up *immensely*. Neat little thing, too, smaller than I was expecting - it's amazing how compact they are when they don't have the massive flatbed attached.

 

Ooh, was nearly very upset then. Being the massive autist that I am, I was pixel-peeking the scan... and discovered this:



A vertical stripe. Went back through the whole scan: it's there on all of 'em, from the upper scan head. Gave it a good wipe with a cloth, rescanned a page... still there. Dead pixel in the scanning element? Disaster!

In a last-ditch attempt to not have to try to explain to the seller why I want to return a scanner with a ONE PIXEL fault... I threw a wet lens-cleaning wipe at the problem.



STRIPE BE GONE!

(No, I'm not rescanning Issue 90, sod that. It's one pixel wide, I'm not *that* anal.)

 

Yeah, there's definitely a difference in brightness and saturation between the two.

Issue 90's cover on the flatbed of the old scanner:



Bright reds, though some parts are blurry where the cover's not quite lying flat.

Same cover, new scanner:

'

Very dark, faded reds, but everything is equally sharp - slightly blurrier than the flatbed at its sharpest, but sharper than the flatbed at its blurriest. Next one I slice up I'll scan with various brightness/contrast settings in Simple Scan until I find the best values. (I say values, there are no numbers - it's just a-bit-more or a-bit-less...)

EDIT:
Easily salvaged, though - here's the new-scanner version but processed with a quick ImageMagick "parallel convert -modulate 110,150,100 {} {.}-tweaked.png ::: *png":



Close enough!

 

I think it's a combination of too dark (it was *really* dark at the default settings, and Simple Scan doesn't track brightness/contrast settings per-scanner so my old-scanner version is brighter than I'd normally have it) and too unsaturated. I can fix the former in Simple Scan (and maybe a bit of the latter if I fiddle with contrast as well as brightness), and then batch-boost the saturation in ImageMagick post-scanning - see the edit above.

 

Decided to install the official Fujitsu/Ricoh driver for the scanner (well, the family of scanners.) Amazingly, there is one - a .deb, even.

The good: comes with a bunch of command-line tools *and* it exposes a bunch of missing settings. Though not the promised deskew.

The bad: it enforces JPEG compression. As in, the data is compressed on the scanner as JPEG before it's sent over-the-wire, then the driver decompresses it(!) and sends it to the requesting application as raw image data. And, sadly, it's noticeable. Now, you can change the compression level or turn it off entirely when using the command-line app or XSane, which is nice... but you can't turn it off when using Simple Scan. Which is what I use.

If I uninstall the official driver and use what I was using before - i.e. whatever's built into the sane database - it does *not* appear to compress the data, or if it does it's much less noticeable. Weirdly, the performance is the same either way - though, granted, I'm scanning at 300dpi, maybe it's different at 600dpi.

Fujitsu's driver, default compression:



Enlarged, as they say, to show texture. Peep the clear JPEG block artefacts visible in the thing. Now, okay, sure, we're at, like 300% zoom, and I'm going to be compressing it myself anyway - but that's the point, I need to start with an uncompressed file or it's going to get compressed twice.

Sane's driver, whatever the default is for that:



Suddenly the halftoning is visible! No JPEG blocks! Lovely!

So, to cut a long story short: I installed a driver then almost immediately uninstalled it.

 

Given Simple Scan's simplicity, the only image adjustments I can do during scanning are brightness and contrast. And what I'm getting out of the scanner is:



Again, enlarged for texture.

My options are to switch to more complex scanning software, or... just batch-process in ImageMagick afterwards. It's a one-line command, and'll probably take a couple of minutes tops?

Here's the same scan as above, only after -modulate 100,150,100 has been applied (so a 50% boost to saturation, everything else left as-is):



It's not exactly what I've got sat in front of me, but it's pretty close. Good enough, I reckon!

 

Finally got the guillotine, but I haven't had time for a proper play. Tested it with an old Cosmo magazine: it took the spine off clean as you like, though it was angled just like when I'm doing it by hand. Might have been my fault, there might be something I can do differently next time. Hell of a lot quicker than the ruler-and-knife method!

I also did a little quick colour processing of a few pages as a test. Best results seem to come from "parallel --bar convert -contrast -modulate 100,120,100 {} tweak/{} ::: *png", which I'm putting here for future reference. Colour accuracy isn't great - one of the pages had a background which came out quite differently - but it's Good Enough. It's a document scanner, not aimed at artists or designers.

Looking forward to seeing how quickly I can get a mag sliced and scanned now!