Scanning the Ookla Way

You may know of different ways to scan. This is how I do it, and I expect similar results from anyone who wants to help scan.

Note on the screenshots in this guide: some are taken from the Mac OS X version of Photoshop, and some are taken from the Classic Mac OS version of Photoshop. Some screenshots will thus be uglier than others (and you may hate all of them).

Preparation and methodology

Preparing the scanner bed

First, make sure the scanner doesn't cut anything off. If you place the book in the corner of the scanner bed, you may get fractions of a centimeter that the scanner undesirably crops off automatically. So you'll want to move the book more toward the center. But this makes it easy to place it crookedly. The way to keep things straight is by placing a guide on the scanner bed. I use a piece of paper with a hole cut from the center of it. The paper is taped to two edges of the scanner plastic (but never to the glass). With a standard 8.5x11 sheet of paper, leaving margins of 2cm makes a practically perfect space for a tankouban to lie horizontally. 2cm-wide strips of thicker plastic would work even better, but I have no idea where I might find something like that.

Here's a scan of the paper template taped to the scanner bed. [Note: I stopped using this eventually. But it may still help you.]

The goal of this is to get scans straight and with nothing cut off at the edges. If you have another way to do that that will allow you to scan two pages at once (see vertical vs. horizontal below, as well), go ahead. Some scanners may cut off nothing at all.

Also, when I scan, I remove the lid from the scanner completely. Closing the lid during scanning will just make it more likely for the book to go crooked. The lid just gets in the way. If your scanner's lid isn't MEANT to come off completely, don't remove it, of course.

Preparing the book

After you've prepared the scanner physically, prepare the book physically. Some people cut apart/debind their manga. I never have. [Note: Not too long after writing this I caved and started debinding. It gives you better results, if you care. I started buying multiple copies of the books so I'd have one to keep. A used manga store can be a big help.] If you want to do this it's up to you. I personally like keeping the manga in book form. I've got a couple simple tips to keeping your manga in good shape, no bent spine or anything like that, while still getting good scans.

To avoid major damage (bent spine), I recommend deliberate minor damage. The key is to make the cover open all the way so that when you turn just the cover and 1 or 2 pages and put the book face down, it lies completely flat.

Here is how the covers usually want to open, and what you want to avoid. Here is what a cover looks like opening all the way so it will lie flat, and what you want it to look like.

To get this result, stand the book up on its spine and open the cover. Keep one page, the blank or copyright info page, with the cover. Use your fingertips in the gutter of the book to separate the page from the next page. A small amount of glue will be adhering the pages together, and stopping it from doing that is the minor damage to inflict.

Once that's done, lay the book face-down with the cover + 1 page open, and make sure it goes flat without adding a crease to the cover as in the BAD image above.

Press down evenly

When you scan it's best to press down on the book so that the texture doesn't show up too much and for various other reasons. But pressing down with your hand can give uneven pressure, and if you misjudge your strength you can damage the scanner. I stack books on top of the manga. Typically I first use two dictionary-sized books and then some nice thick hardbacks on top of that. Textbooks are also ideal. They should press down evenly on the one page you're scanning, or if you can manage two at once, you'll need an equal number for each side.

Scan in order

The next thing to try to do is to scan the whole book at once, or to scan the first half in one sitting, starting from the first page and going to the back, and the second half in a different sitting, starting from the back and going toward the front. Why? The book will be more pliable this way. If you start scanning from the middle, you run a much higher risk of damaging the spine.

Vertical or horizontal? What about gutter shadows?

Here's a more-or-less actual-size scan of the book lying horizontally. Notice that at the top there's a shadow on the first page and a brighter area on the second page. Also notice that the left and right edges of the second page have curved lines. These are reasons not to scan with the book horizontal, though I have done it this way in the past.

But Here's a scan of the book in vertical position. There's no curved lines. Very little shadow comparatively, and no brightening. If you look closely you'll see that there's a bit of lateral image compression toward the gutter, but this is pretty minor, and the flatter the page lies the less of this you'll get. I'd recommend going with the vertical scan at this time. [Note: While vertical scanning doesn't curve the lines in the gutter, it may cause blurriness where the paper isn't lying completely flat. This is a reason that debinding will give best results.]

Scan specifications?

Scan in Greyscale, 8-bit or 16-bit mode. 16-bit may help with very dark or very light tones, but after you get it into photoshop change it to 8-bit. 16-bit is just too large as far as filesize goes. I'm not even dealing with color scans in this guide.

What resolution? Generally we've done 300 dpi in the past, but 360 dpi will give better results; it's a more exact multiple of the LPI used in the tone screens in tankouban-sized manga. It will also make a larger filesize, but that consideration will be dealt with later. [Note: As of 2003, we were usually scanning at 600 dpi, and it appears that quality at 360 can be at times worse than at 300 depending on what kind of scanner you use. (The lpi explanation I used may be bunk.) YMMV. Experiment for what works best for you. We generally found that scanning at 600 did the best job keeping details. Calibration discussed below still applies though, and doesn't really vary by what size you use. If you are scanning for professional work at one of the manga companies, use 1200 dpi bitmaps. ;)]

Scanner Calibration

Now I'm going to explain how to get scans looking good. Instead of just trusting the scanner to do the best it can and then just adjusting each individual page afterward in Photoshop or counting on the editors to get the page looking like the book, it's possible to calibrate your scanner settings to give you a good picture every time. Unless your scanner software is horrible.

Starting the TWAIN interface

In order to calibrate the scanner, I use Photoshop and TWAIN scanning. If you don't have Photoshop you'll have to adjust these instructions to fit your situation.

TWAIN can be found in the Import submenu of Photoshop's File menu. If you do not see it in your menu, see if any other option points to something that looks like your scanner. If there's nothing there, check your scanner manual for how to use TWAIN.

Advanced mode

With my scanner, a UMAX Astra 1220U (which I would NOT recommend you purchase, but it does the job for me [Note: I later got a Canon LiDE 80, which is excellent. There are probably better ones out there now.]) the TWAIN interface is called VistaScan. This is what the basic screen looks like. Your scanner probably has a similar, if not as ugly-looking, basic interface. We do NOT want to use the basic interface. So I click on the Advanced button instead. Other scanning programs usually have a similar distinction.

Select the test page

In order to calibrate the scanner, you'll need to find a sample manga page that has three basic things: one or more large areas of black, one or more large areas of white, and areas of both dark and light tones. For this guide I'll use page 72 of Black Cat vol. 3 as an example page.

Automatic settings?

Most scanners will have an option to try to select the best settings automatically. I'll show why that doesn't do a good enough job. First I put the manga on the scanner, with books stacked appropriately. Notice that the "Auto Adjustment" box is checkmarked. I set the dpi to 360. (8-bit Greyscale is called "B/W,Photo" by VistaScan.)

Pressing the Preview button will scan the whole scanner bed and make it appear on the left. It will also give the program a reference to use for the auto adjustment. (Even if you don't try out the auto adjustment to see how poor it is, you do need to use the Preview function at least once to get the right page area indication up on the left side here.)

Set scan area

The previewed scan on the left will have a selected area to indicate what to scan in the full scan. Shrink the selection to the page area, like this. For the calibration scan, the selection should pretty much cut off all non-page areas. When you go into production scanning, it's best to leave a bit of margin for error around the edges (in case the book is slighly moved) so nothing gets cut off that wasn't supposed to be. The editors can crop this off later. But for the calibration, pretty much crop it.

More numerical settings

Besides the place to adjust dpi and color/bit depth, the scanner program should have a place to adjust brightness, contrast, highlights, shadows, and midtones and/or gamma. Usually these options are five, but some programs such as Canon's separate them into six, midtones and gamma being separated somehow.

Here is found this section of Vistascan, by clicking on the button below "Advanced." From the top down, the five boxes and sliders are brightness, contrast, highlights, shadows, and gamma. Brightness and contrast use positive and negative numbers. Highlights and shadows use values from 0 to 255. Gamma uses numbers from 0 to 10 but including neither. The defaults are brightness 0, contrast 0, highlights 255, shadows 0, gamma 1.0. (Though usually the scanner program will use something other than 1.0 for gamma; 1.2 is common.)

Scangear CSU from Canon (picture taken from a PDF) looks a bit different, but it has the same options. This screen shows only shadows and highlights (black point and white point) and midpoint. The button to the left of the "mountain range" button reveals the gamma setting screen, and the one to the left of that has brightness and contrast. (For Canon users I recommend ignoring the "Mid-point" setting and using the gamma setting in the other screen. Both can do essentially the same thing, but the "Mid-point" setting uses a value from 0 to 255 and the gamma setting uses the "1.0" etc. value as used in Photoshop and as discussed in this guide.)[Note: Once I got a Canon, I found it was actually more convenient to use the midpoint setting. If you know what you're doing, you can experiment for the best results for your own setup.]

Shadows? Highlights? Gamma? Huh? Why?

Just what are these settings anyway, and why use them? I don't use Brightness and Contrast because they are very imprecise. Using the Shadows, Gamma, and Highlights settings allows you to precisely control how the image scanned from the paper will appear when it goes on screen.

A scanner effectively gathers light from the scanned surface one line at a time. If you're scanning at 360 dpi and the book is about 4.5 inches wide and 7 inches tall, that's 1620 separate pixels read across the page. For each of the 2520 times those 2620 pixels area read by the scanner, each pixel contains a number value from 0 to 255. That number corresponds to the brightness of the gathered light for that part of the book on the scanner: a value of 0 means no light was gathered, indicating the image there is pure black, and 255 means the maximum possible light was gathered, indicating the image there is pure white. Ideally, because manga is printed only with black ink, all pixels would have values of only 0 or 255 (if the resolution was high enough, like perhaps 2880dpi, but various factors keep this from happening. First off, manga paper is not stark white. It's more of a warm color. It also has texture, and the light from the scanner head will cause minor shadows in the paper grain which reveal this texture. The black ink still reflects some light, so a value of 0 from it isn't possible either. Also, the scanner itself just won't give perfect results due to technological limitations.

For this reason, the scanner settings for a good scanner will allow you to tell the scanner what values to use for Shadows, Highlights, and Gamma. A Shadow value of 25 tells the scanner "take all the pixels with luminosity 25 or less and assign them a luminosity of 0." A Highlights value of, say, 220 tells the scanner "take all the pixels with luminosity 220 or above and assign them a luminosity of 255." What happens to the pixels in the range between those two values (25 and 220 in this example)? They get distributed from 0 to 255. What the gamma number does is tell the scanner how evenly to distribute the middle pixels: a gamma of 1.0 means they get distributed evenly, a gamma in the range 1.0 to 10.0 means they get bunched more toward the white end, and a gamma in the range 0.0 to 1.0 means they get bunched more toward the black end. By setting these three values then one should be able to get a scanned image that looks like what's on the page.

Trying Automatic Settings

A good scanner program will have an option to attempt to automatically pick the best values for shadows, highlights, and gamma, but it generally won't do a perfect job. Let's see how well my scanner does.

Now that I have my test page on the scanner bed, with my scan area set to the page, I make sure "Auto Adjustment" is selected, and then I press the Preview button. It makes a quick scan of the image, and then the automatic settings appear. VistaScan has chosen the values Shadows 0, Highlights 232, and Gamma 1.6. Now I test these settings out by hitting Scan.

Here's the raw scan with automatic settings, saved in full glory as a 2.4 MB PNG file.

Now in Photoshop I take a look at the result (open it up and take a look yourself). How well did it do at picking the numbers? Not that well. White doesn't look very white, and black doesn't look very black.

The Histogram

To see just how far off it is, bring up the Histogram. Here's what the histogram looks like for this image.

The histogram shows a graphical representation of the luminosities of all the pixels in the image. From left to right are luminosity levels 0 to 255, and the vertical axis represents number of pixels at each luminosity level.

Histograms of manga scans will have two things that stand out: a "mountain" on the left and a "mountain' on the right, as in this image. The one on the left, in this picture centered at 50, represents the pixels that are supposed to be black, and the one on the right, in this picture centered at 243, represents the pixels that are supposed to be white. In a perfect scan there would be no mountain at all, just huge numbers of pixels at 0 and 255 with varying amounts in the middle (pattern depending on the types of tones on the page).

The histogram confirms what you can see with the eye: black isn't black and white isn't white. The histogram tells you how much they aren't white and black.

Obviously, the automatic settings didn't do a very good job. But isn't it good enough? Why not just scan with automatic settings, and then let the editor adjust levels later, or perhaps adjust the levels yourself and record an action to do so automatically after every scan? I'll mention three reasons.

First is that the automatic settings will choose different values depending on how much black and white and tones are on the manga page, and different manga pages obviously have different pictures and thus different amounts of black ink.

Second is that the person assigned to edit often won't own a physical copy of the book and thus can't check what's on his screen against the book while doing Levels adjustments, something that is particularly meaningful when dealing with midtones; without having a copy of the book it's very difficult to get tones the right relative shade just by guessing from the scan.

The third consideration is colorspace resolution. I don't mean dots per inch in this case, but number of levels from black to white. If the editor plugs in a Shadows number of 50 and a Highlights number of 210, that means 50 levels are being chopped off the left side and 45 levels are being chopped off the right side. Instead of having an image with a full 256 greyscale values, now it's only got 161 of those 256 values that are used. This means loss of detail, and it's the tones that generally will suffer. If you instead plug those numbers into the scanner settings, the scanner will give you an image with correct black and white values but utilizing a full range of midtones.

Now let's really get down to business.

Scan your test page with zeroed settings

It's time to start the calibration of the scanner now, if you're using this guide as step-by-step instruction, now that most of the "whys" are out of the way. I recommend reading through this twice: one time to open up my scans as I talk about them and try the same adjustments on them, and a second time to follow along while scanning your own test page, following the same steps in order to calibrate your own scanner.

My selected test page on the scanner bed, I go back to the scan program and turn Auto Adjustment off. Then I set Brightness to 0, Contrast to 0, Highlights to 255, Shadows to 0, and Gamma to 1.0, like so. And now I press the scan button.

Here's the raw scan with zeroed settings, saved in full glory as a 2.5 MB PNG file.

Save the scan now if you're scanning your own test image. You'll need it later. (Just save as PSD for convenience and accuracy in this case.)

Why reset everything to the zeroed values, if we know the numbers should be changed at least as much as the automatic settings changed them? Because if we leave them at the automatic settings, once we get some numbers back out of Photoshop, we'll have to multiply those numbers by annoying conversion factors in order to get the right new numbers to type into the scanner settings, and it's much simpler and more straightforward to keep the numbers in an exact correspondence. Multiplying and dividing by fractions of 255 isn't very fun either.

Histogram of the unadjusted zeroed scan

Now I open up the Histogram in Photoshop again. I recommend you get the unadjusted image and do the same. It looks like this. You should be able to see the black peak at 22 and the white peak at 213. Your own scan will of course have peaks in different places.

Finding candidate values

I'll now find good test values to use for the scanner settings.

First, the black. After rotating the scan, I go to one of the large black areas and select it. Using a noncontiguous selection in this step can yeild odd results in the histogram for one reason or another, so I stick to one connected area. After selecting the area, I bring up the histogram again. You'll see that only the selected pixels are used to determine what's in the histogram, so only the black peak is visible.

All the pixels in the black peak are supposed to have a value of 0, so I focus on the right side of the black peak. I pick a spot about 1/3 of the way up from bottom to visible top of the peak, and write down the Level value for that spot, in this case, 27. That will be the candidate Shadows value.

It's important to write the numbers down in a place where they won't be lost so that you can have them for future reference. My scanner saves the settings of the last scanned image, but not all scanner programs will; check your scan program documentation to see if there's a way to turn this option on if it appears to be off.

Next, the white. Similarly, I select a white area and look at its histogram. As this peak is wider, I focus on a point about 1/4 of the way up the peak on the left side, in this case level 198, and write the number down.

Selecting a gamma value

Now that I have my test black and white points, I need to find a gamma value. I do this by going to Levels and scrolling to a tone area on the page, because Levels is where you set gamma, and tones are the areas where gamma's affect is most obvious.

Once in Levels, I input the initial values of 27 and 198 that I wrote down, like so. Notice that when typing in the Blacks value in the top left box, the black arrow moves to the right, and when typing in the Whites value in the upper right box, the white arrow moves to the left. The grey arrow, the Gamma arrow, stays equidistant from the other two arrows. Now it's time to move the gamma arrow.

For this part, it's necessary to remove the book from the scanner so you can compare what you see on the screen to what's in the book. Just as I've moved the screen to the part with tones, I look at the tones on the page. Comparing them visually I can see that the tones on screen are much darker than the ones on the page, especially the dark grey which is almost black onscreen and not near that dark on paper.

Tones too dark means the gamma value needs to increase; the gamma arrow needs to be moved to the left. Think of it this way: the pixels right above the grey arrow will be given a 50% grey value, so the closer to black the arrow is moved, the fewer pixels will have less than a 50% luminosity. In all cases and scanner models I've seen so far, scanner calibration requires a gamma moved to the left.

So I use the mouse to move the grey arrow to the left (being sure to have "Preview" checkmarked) until the tones onscreen match closely the perceived grey levels of the tones on the paper. In this case, a 2.4 Gamma looks just about right. Because of VistaScans's limitations, I stick with a precision in the tenths digit rather than the hundredths digit; if your scanner settings can use hundredths, there's no need to round it off. I write down 2.4.

Flipping back and forth between these two pictures you should be able to tell what difference the gamma makes. Pay attention to the hair highlights as well as to the tone areas. They're much more visible now.

Note for Macintosh users: It's a good idea to set your monitor gamma (in the Monitors System Preference / Control Panel) to the PC and web-graphics default of 1.8 instead of leaving at Mac-default 2.1, while you're working with scanning and editing manga. I currently have Photoshop's colorcorrection turned completely off as well. Otherwise, final image gamma may be a bit off.

Rescanning using candidate values

Now that I've written down the values 27, 198, and 2.4, it's time to rescan. I input the numbers into the scanner settings and hit Scan. Then once in Photoshop I rotate the image.

Here is the raw scan with candidate settings, saved in full glory as a 1.6 MB PNG file.

Evaluating the rescan

Now the question is whether the settings are good enough. Visually, there's still grey in white areas. The histogram of the new image shows there's still a significant distribution to the pixels that should be white on the right, and a bit still to the black pixels to the left. What does this mean? The guessed values weren't quite good enough. I'll have to raise the Shadows value a bit and decrease the Highlights value more significantly.

Back to the unadjusted scan

Now it's a process of trial and error. I go back to the unadjusted scan that I saved earlier. To help me get a better value, I select the whole page with the marquee (or Select All) and then Exclude from the selection questionable areas including the very-slightly shadowed area near the gutter and the non-book area from the scanner bed.

Now I look at the histogram again. First I focus on the black side to try to find the turning point in the curve, paying attention to pixel count. Moving right from the 27 value, pixel count keeps dropping sharply. Level 29 has 19253 pixels and Level 30 has 15851 pixels, a drop of almost 25%. But then level 31 has 14215 pixels, a much slighter dropoff than before. 30 I'll call the turning point, so I pick 30 as my new test black value and write it down.

Now, to find a white value, instead of looking foor a turning point, I'm going to find the level with a similar pixel count to the black value I just found. This turns out to be Level 189 with a count of 15839. I write down 189.

Now the gamma. I could manually set it like before, but there's something else I can do for that. Going back to levels, I set Shadows, Gamma, Highlights to the old values of 27, 2.4, 198. Then I position the mouse pointer directly over the grey arrow. I use the keyboard tab to get to the Black box where I type in 30. The black arrow moves, and the grey arrow moves accordingly. Then I click the mouse button without moving the mouse and the grey arrow jumps back to the same spot, giving it the new value of 2.52. Now I do the same with white: tab to it and type in 189, causing the white and grey arrows to move. Then I click again with the mouse, and the grey arrow jumps back--to 2.38. If I round it off, it's 2.4. Looking at the tone part of the image, it looks right. So 2.4 is the number I will use this time again.

This clicking trick with the gamma seems to work pretty well for getting gamma in a good spot when moving white and black values. Now I've got my three new values so it's time to rescan again.

Rescan again and evaluate

I put in the settings and scan again, and this is what I get:

Here is the new raw scan with the second candidate calibrated settings, saved in full glory as a 1.4 MB PNG file.

You may notice that the filesize keeps going down. As more pixels that should be white with a value of 255 actually get that value, and should-be black pixels get 0 values, there are more areas of true solid color in the image, which makes it more compressible, causing filesize to drop. That's exactly what you want, as long as nothing that shouldn't be black or white becomes that way.

The histogram of the new scan looks like this. Black looks pretty good, but white still has a bit of a slope. However, this is probably as good as I'm going to get on the white end without damaging the light grey tones. So I'm going to stick with these numbers and do some final adjustments using curves instead of messing with levels any more.

So my final numbers are Shadows 30, Highlights 189, and Gamma 2.4. You'll get different numbers; write your numbers down if you haven't already!

The scanner is now calibrated, but before zipping it up to send to the editor, there's a couple more steps we'll do.

[Unfortunately, at this point my html file got corrupted, and a whole section on using curves to zero out the ends to get rid of dirt specks was lost. I don't feel like recreating it, but here are the relevant screenshots:


I hope you can figure out from the filenames and the pictures what the heck I was trying to do. ;)

Then I talk a bit about making an action to do the curves part automatically for you whenever I want, but only the last part of that survived. Sorry.]

click on Create New Action. In the window that pops up, I give my action a name and assign it to some unallocated function key. Once I hit Record, Photoshop starts recording everything I do.

So I do the whole Curves action again: bring up Curves and draw the two little lines. Once this is done, I hit Stop Recording (the square button) on the Actions palette.

The action is now recorded. If I expand the view in the palette, it looks something like this. To use this action, after every scan I simply press the F9 button on the keyboard. Now it's time to save the scan.

Saving as PNG

Rule 1 of manga scanning and editing: if the picture isn't ready to be downloaded from the website, never save as JPEG. That means the person scanning will never, ever save as JPEG. (Rule 2 is that only color pictures are ever considered to be saved as JPEGs, but that's only to be done by the editor and thus has no impact on someone scanning.)

Always save in a lossless format, meaning that when you save it and open it back up, every single pixel is the same as it was before you saved it (this is impossible with JPEG, which introduces compression artifacts even at the highest quality). There are two obvious choices: PSD and PNG. Saving as PSD is the default in Photoshop and a good choice. However, it's nice to try to keep raw scans of a 20-page chapter around 20 MB or less. Especially if you scan at 360 dpi, PSDs will often be larger in filesize than desired. PNGs are smaller and slightly more difficult to do. But they're not TOO difficult once you get the hang of it. So let's do PNGs if chapter size is above 20MB.

To save as a PNG, I select Save As from the File menu. In the window that pops up, I select PNG from the Format drop-down menu. (Where to save the picture depends on circumstances, but generally all scans of a chapter should be in the same folder; this should be obvious.)

After I hit Save, a PNG Options window pops up. The Interlace option to use is None. (Using Interlaced will increase the filesize substantially, with no benefit for this kind of work.). After hitting OK it's saved and I can go to the next page to scan.

But that's not the only step in saving PNGs: in order to get small files (and fix a Photoshop bug), it's necessary to do one more thing to the PNGs: crush them.

Sidenote to Mac users: [As of 2005, this paragraph is pretty useless unless you have an old version of Photoshop. See below.] Photoshop's built-in PNG plugin (on mac and PC) isn't very good, though it does the job. A much better option is SuperPNG from fnordware, which creates much smaller files, gives you more options (including allowing 16-bit PNGs), and fixes the gamma bug (if you uncheck "save metadata"). I'd recommend this for PC users too, but the PC version is a bit behind and won't handle indexed PNGs--which doesn't matter too much for scanning, but as all final-version TW chapters are saved as indexed PNGs that could create problems. Hopefully the PC version will be at 1.0 before too long.

Using pngcrush

[Note: As of 2005, the PNG universe is quite different from our awareness of it in 2002. Photoshop CS+'s built-in PNG handling is much better than PS 7's was, and there are also better options than pngcrush available, such as OptiPNG, PNGCrusher (for Mac OS X), and AdvanceCOMP's advpng utility. If you are confused already, Photoshop CS's PNG handling is probably fine for you. But if you want to wring all the bytes possible out of your PNGs, explore all these options thoroughly. The following is the old information, with slight variation, so make of it what you will.]

pngcrush is a small command-line program for batch-processing your saved pngs in order to optimize filesize without changing image date. This is possible because there are many different ways to compress data within the PNG file format; pngcrush find the best one for each picture using trial and error. It's thus a bit slow for very large pictures, but it doesn't require you do anything other than wait for the program to finish its job.

The pngcrush homepage is at this link. If you go to the download page you may be confused and download the wrong thing. Since it's an open-source program the source code is available and prominent on the page. The actual program is listed further down the page under pngcrush-executables. As of 2005 the most recent PC (DOS) version is 1.5.10 and you will find it at this link. (I don't really know what the win32 version does.) If you are on Mac OS X for it. [Note: There used to be an easy location, but now your best bet may be installing commandline tools packages like Fink. Sorry. Try PNGCrusher above for ease of use.]

You're probably wondering why I'm saying to use a command-line program on pictures. Yes, it doesn't seem to make that much sense, and it's not incredibly user-friendly. Someone could easily make a graphical front-end for the program, but I haven't seen one around. Still, for a command-line program, it's not hard to use, and the built-in help is fairly well explained. But so you don't have to muck around with it, I'm going to explain how to use it.

Navigate to the folder where your output pngs are (it's easiest to do this once all pages of the chapter have been saved). Make a new folder there called "optimized" or "opt" or whatever you want to call it. Still in the folder where your pngs are, run pngcrush with this commandline:

pngcrush -dir opt -rem alla -m 4 *.png

This tells the program to put the output pngs into the "opt" folder (put here whatever you named your folder). It also tells it to remove extra bits that Photoshop adds to the png (including gamma information that Photoshop will likely write incorrectly, making the png too dark compared to a jpg; after running pngcrush the png won't be too dark any more). And it tells it to use method number 4. pngcrush then does both things (remove the extra info and recompress according to method 4) to all the pngs in the folder.

For every well-adjusted greyscale manga image I've seen, method 4 has been the one that results in the smallest file. If you really want to be sure you get the smallest png possible, use "-m 0" and it will try all 10 standard pngcrush methods. This will obviously make the pngcrushing take longer. If you really really want to make sure, instead of using "-m 0" use the "-brute" option. This makes pngcrush go through its 130 or so other methods, using all allowable permutations of the three parameters. The 10 main methods are included in this; method 119 is the same as method 4, for example (if I recall correctly).

Here's a screenshot of the program at work.

As a comparison, Here's a screenshot of the relative sizes of three pictures: the uncrushed scan using the second settings, the uncrushed scan after curves, and the crushed final version. Curving and crushing knocked 36% off the filesize.

Here is the final raw scan, curved and crushed, saved in full glory as a 1.0 MB PNG file.

Zip 'em up!

Hopefully you know how to use WinZip on PC or DropZip (or whatever else works for you) on mac. Zip up the chapter's scans and that's pretty much it. Well, do make sure you're zipping the crushed scans from the "opt" (or whatever) folder, not the uncrushed scans).

It's a good idea to keep the crushed scans until at least a week after the chapter's public release, if not longer.

Final considerations

This guide is not the Ten Commandments. Most rules are not hard and fast. If you find a better way to do somehing, start using it! Just make sure it only improves the process without worsening any component of it. I keep finding better ways to do things.

While making this guide I noticed that I can get rid of the shadow in the bottom left corner of the final scan by cutting a notch in the paper template on the scanner bed, so that there's nothing to cast a shadow in that spot. If you don't use a paper guide, that wouldn't be possible; the edge of the scan bed itself would then cast a shadow that you couldn't get rid of without moving the book.

I may also be able to rework things so I can scan two pages at once with the current vertical orientation by narrowing the width of the paper guide. Maybe; it will be a close fit. That would be ideal since scanning would take less time. I'm not exactly clear on where the scannable image cutoff is; I'll have to figure that out.

That's pretty much it for now. Feedback is basically useless at this late date. :)

Ookla The Mok

First draft completed 13 September 2002.
Some updates, though I still didn't fix the corrupted part completely, 12 November 2005