The layout for this guide assumes that you have at least a screen resolution of 1024x768. If you don't, it probably looks a little funny. Sorry about that.
I am always interested in receiving suggestions on how to improve this guide. You can contact me at: fywp_cca@hotmail.com
1. Preparation
2. Scanning the page
3. Image alignment
4. Cropping the image
5. Image cleanup
6. Saving the image
7. Multi-page images
8. Black & White Image Cleanup
9. Storing B&W Images
Appendix A. CCA Picture Format Standards
Appendix B. Example cleanup steps
However, since there is no real need to do exact color matching, it's sufficient to just have your monitor adjusted properly for brightness and contrast.
Turn the lights off and draw any curtains shut. Display a black image. Set contrast to 50%. Set brightness all the way down. The screen should be completely black. Now start to turn the brightness up. I like to have my brightness set as high as possible -- right at the point where black becomes visibly grey. Now display an image with large areas of black, white, and color. Adjust contrast so that white is bright white without bleeding. Same should go for colors -- bright colors should be vibrant but not bleeding. Dark colors should still be visible and not muted into black.
Don't sweat the adjustment too much. Unless your monitor is burned out & total crap, as long as your scans look ok on your monitor, the rest of us should be ok.
Many scanners will allow you to set a light & darkness range for a scanned image, through the use of the histogram tool in your scanner driver (sometimes it is named differently). What the histogram tool setting does for the scanner is take that 30 or 36 bit scan capability and stretch a selected light/dark band into the 24 bit range the computer uses for a better looking color distribution. If you don't understand that, don't worry about it -- many newer scanners automatically do this adjustment for you. If your scanner doesn't do this automatically, I recommend you NOT try to use this type of tool! Again, the color tone range on most comics is fairly limited (even those fancy Image comics), and unless you've got an automatic adjustment, to properly use a histogram tool, you need to adjust it for every single page or run the risk of screwing up light & dark regions.
An easier solution is to mark an area where you will scan, make sure the page you are scanning falls there, and then rotate the the page image straight via software.
If you simply press the comic to the glass and scan, often in the result you will see a "ghost" of the the other side of the sheet of paper. Scanners use a very bright light which is strong enough to penetrate through the paper, and the image on the back side is reflected back. This is especially true of comics printed on newsrag.
To avoid this reflection, put a black piece of paper behind the sheet that you are scanning. It won't necessarily cut the reflection completely, but the reflection will not be anywhere near as strong as if you had not done this.
If you scan with the top up, I suggest you also cover the rest of the scanner glass with dark paper to prevent ambiant light from interfering with the scanner. You can keep the comicbook pressed flat by placing a thick magazine or heavy book over the while thing. (I place a relatively heavy software box along the spine to keep that pressed down, and a couple paperbacks to hold the outer edge down.) Be careful not to scratch the glass/plastic with the staples that hold the comic together!
Now that you have your scanned image, you may
find that the page does not look straight, as in the image to the left.
It is important to remember that in many cases, it is not going to be possible
to scan a page image as a perfect rectangle. Many comic books are
not cut with pages at perfect 90 degree angles! Plus, unless you
actually unbind the comic, the curve of the pages will tend to prevent
the page image from coming out as a perfect rectangle.
By the way, notice how the area around the comic is black? That is because I backed the page with black paper, like I said in step 2. |
rotate pic 1 line running vertically |
rotate pic 2 line runs along page edge delta x/delta y = 28/2850 = 0.009824561403509 inverse sin(0.009824561403509) = 0.5629149598494 Rotate 0.56 degrees, rotated counter clockwise |
We want to get the image as straight as possible.
First you must decide what will be considered to be the 90 degree plain.
If in question, I almost always pick the vertical outside edge of the page.
If there is a large border margin for the page, and the artist was capable
of putting the panel frames in straight (some artists cannot), I will choose
that border edge. Otherwise, you can probably pick a line of text
since text is almost always ment to run perfectly horizontal.
In this example, I'm going to use the left edge. In Photoshop the command to rotate at a particular angle is Image -> Rotate -> Arbitrary (Alt-I, E, A). (see a pic of menu) To pick your angle, you either have to have a very very keen eye (and be able to measure in hundredths) or... you can be like me and use a calculator. I suggest you zoom to at least "actual pixels size" (ctrl-alt-0) when you do this for better precision. Make sure you have the Info window turned on (Window -> Show Info) and displaying measurements in pixels (File -> Preferences -> Units & Rulers -> Ruler). Select the pencil tool (N). Click (and hold) one point along the edge that is to run straight. Now drag to another point on that same edge. You see delta-X and delta-Y on the info box as you move the crosshair? You will also see that it shows you an angle, but since this is measured only in tenths, this is not precise enough for us. Ok you're holding your line running along the edge you want to rotate to a 90 degree? Remember the delta-X and delta-Y. If you picked the outside edge, X will probably be much smaller than Y. Use the calculator program that comes with Windows (Start button -> Run -> Calc). Make sure you put this in scientific mode (see pic), and divide delta-X by delta-Y. Now we want to take the arc sin of this to get the angle of rotation of this line to the vertical. So hit "inv" button on calc, then "sin" button. (Or "i", "s"). BTW... If you come up with a POSITIVE number, this means you should rotate clockwise. If you come up with a NEGATIVE number, this means you should rotate counter-clockwise. Take the resulting number and use this as your rotate angle in Photoshop. Now your image should be rotated straight!
|
Dashed line is path of polygon lasso. Unlike crop, poly lasso lets you select the page without chopping off any of the page border! |
Ok, now that you have your image straight, you want to get rid of anything
that is not actual page. So, there are two tools to do this: Crop
(C) and Polygon Lasso (L, in polygon mode instead of the normal freehand
mode). You should pick one or the other.
I usually crop when there's a page border. I use polygon lasso (see pic on left) when the image extends all the way to the edge of a page, or any time when I come close to cutting out part of the actual picture. The crop tool is self-explanatory so I will skip the instructions for that. To use polygon lasso, you will click at the four corners of the page, in order. This follows the exact edge of the page, even if it is not a perfect rectangle. Once I have closed the lasso, use Edit -> Copy (or ctrl-c) to copy the lassoed area to the clipboard. Then File -> New to create a new image (or ctrl-n). By default Photoshop will have the size defined to exactly fit any image stored in the clipboard. Very handy! You can create the background of the new image as "white", "background" or "transparent". I almost always use "background" with the background color set to black. The only exception to this is if most of the page edge is white, in which case I will either choose white, or I will use the eyedropper to set the background to the whitish color of the page edge. Once your new blank image is created, make it active and Edit -> Paste
(or ctrl-v) to paste your clipboard image to it. Once you do this
you can close the original image.
|
Curves - pic 1 Without curve |
Curves - pic 2 Colors curved to dark |
Curves - pic 3 Colors set curved to light |
Curves - pic 4 Light tones made lighter, dark tones made darker, midrange left alone |
The most important thing you can do with a color image is set light & dark correctly, also sometimes called gamma correction or gamma curve. In Photoshop this is the "curve tool". (There is also a histogram adjustment but this is too crude for what we want.)
To get to the curve tool in Photoshop, use Image -> Adjust -> Curves (I, A, V) or Ctrl-M. Initially (i.e. unmodified, see panel "Curves - pic 1") it looks like a straight line, yes? The X axis represents the current tone bright/dark value. The Y axis represents the "changed to" value.
In "Curves - pic 2": Towards the dark end of the X axis, colors are darkened very visibly, moderately in midrange tones, and not much at all towards the light end.
In "Curves - pic 3": This is basically the reverse of pic 2.
"Curves - pic 4" represents what usually you will actually want. Blacks are very dark, whites are much closer to white than grey, and mid range colors stand out clearly instead of having a slightly washed out look. The actual "severity" of the curve that you use can change a lot from book to book. My rule of thumb: Make blacks as black as possible and whites as white as possible, without losing any detail in the image. I prefer to let black to look dark grey and white to look light grey rather than lose image quality. With the newer books, higher quality printing means that the curves can be quite gentle, just enough to bring out colors more strongly and remove the "hazey" quality that a lot of scanners have.
The one other big advantage of blurring a dither pattern is that it will result in a smaller file with better jpeg compression and fewer color artifacts. In the page used in the above examples, saving with blur resulted in a file about 50KB smaller than without.
Apologies that the table below is so wide, and that the image is a really
poor example due to the limited color palette.
The first (this) row may be a bit confusing. You should probably
look at the pics & explanatory text on this table in order. At
the far right is the ORIGINAL unretouched image, as it came off the scanner.
Pics brown6 and brown7 to the immediate right are to answer the question
"why not just use auto-levels?"
The second row is for de-browning via Color Replace. The third row is for de-browning via Hue/Saturation. |
pic. brown7 Level tool applied to brown2 (brown2 is the the image directly below). Correcting the brown color with Color Replace first helps the auto-level function quite a bit, but this image still has a bluish cast. You can do better by setting the level tool manually or by using the curve tool. Photoshop permits settings for almost all image adjust tools to be saved. Between this and the ability to batch process, there really is no good reason to be using auto-levels! |
pic. brown6 Auto-Levels applied to the original image. I don't particularly like to use the level tool on anything except black & white images, although if you must do channel adjustments under RGB or CMYK, the levels tool is certainly useful for viewing the image histogram. Auto-levels will force the darkest and lightest areas of an image to pure black and white. However, using it to change brown paper (the brown is the lightest color) to white often results in a "dirty" look, as well as doing strange things to the coloring. In addition, in the rare instances where there is no black on the page, auto-leveling will completely destroy the appearance of a page. |
pic. brown1 the original image off the scanner |
For color replace, eyedrop a mid-tone from an uncolored (browned) area of paper. Generally, I set the fuzziness between 30 and 100, and tend to increase lightness between 20 and 50 while reducing saturation just a little bit. You should NOT try to change the paper to bright white -- if you do, you'll just end up destroying all the coloring. I attempt to approximate the original color of the newsrag. |
pic. brown2 Image after the Color Replace to the left is applied to the original image. |
After the color replace, use the curve tool (or levels if you prefer) to whiten & brighten up the top color range of the image a bit. |
pic. brown3 Paper tone is about what it would be for the original comic, and color saturation remains about what it was from the original scan. Coloring for comics printed on newsrag often do take into consideration the affect from being printed on less than pure white paper. I like this appearance just fine, and it's probably what I'd go with for my own scans. |
For Hue/Saturation, change the Edit pulldown from Master to Yellows (or reds, or whatever, it doesn't really seem to make a difference) and then eyedrop a mid-tone from an uncolored (browned) area of paper. This sets up the range bars you see between the color bands. Adjusting these is analogous to the "fuzziness" of color replace. For some reason when using the Hue/Sat control, I tend to reduce saturation more and increase lightness less than when messing around with Color Replace. There's no particular reason for doing it this way. Again, I don't try to change the paper color to bright white. |
pic. brown4 |
After the color replace, use the curve tool (or levels if you prefer) to whiten & brighten up the top color range of the image a bit. Note that this isn't the same curve used directly above, and both this and the desaturation from the Hue/Saturation are the main reasons for the difference between pics brown3 and brown5... |
pic. brown5 About the same as brown3 except yellows aren't saturated as heavily. |
All of these problems may be corrected the same way -- by manually touching up the affected area. In "touchup pic 1" you see the original scan. There's a staple that I want to get rid of. The simplest way to do this is to overwrite the area with another similar area of the picture. I first use the lasso tool (L, in freehand mode) to select an area to repair. The lasso selection doesn't have to be perfect the first time. Photoshop allows addition to the lassoed selection by holding the Shift key, and subtraction from it with the Alt key. In "Touchup pic 2" I've selected an area of staple.
By moving the lasso cursor into the center of the lassoed area, I can drag it to another place on the picture that has an appropriate pattern for the problem area (touchup pic 3). I copy this area (Edit -> Copy or ctrl-c) and then paste and move it over the original lasso point (touchup pic 4)
This process is repeated this until the problem is fixed (touchup pic
5)
Touchup Pic 1 |
Touchup Pic 2 |
Touchup Pic 3 |
Touchup Pic 4 |
Touchup Pic 5 |
In some instances where there really is no efficient way to use this copy/paste approach, I usually use the airbrush.
I suggest that you do touchup after rotating & cropping your image but before applying any color correction (e.g. blur, curve tool, color change) to it.
So that I'm not performing any unnecessary effort, I view the picture at the final resolution to see exactly what spots on the image are in need of correction -- very small problems (such as a spec of dust) will tend to disappear when downsizing the image from 300 dpi to 150 dpi. (For a quick view... ctrl-alt-0 zooms to actual size, followed by ctrl-minus ctrl-minus shows a 1:2 zoom)
(Note: The quality "values" here are for PhotoShop v5.0 -- version 5.5 now has a scale that extends from 0 to 12 instead of 0 to 10.)
As a general tip, since color response for a lot of scanners is not even across the scan bed, scan one page right side up and the other upside down.
I will show an example of this later.
There are two separate styles of black and white images. Greyscale
images, which make use of many gradations of light & dark, and pure
black & white. The type of image affects what you should be willing
to do for cleanup.
This is a greyscale image. |
This is a black & white image. |
It should be pretty apparent what type of image you're looking at, but
let's bring up the Levels tool (Image -> Adjust -> Levels, or Ctrl-L).
In other programs (including I believe PaintShop Pro) this is also referred
to as a Histogram Tool. The histogram is a representation of the
number of pixels distributed in each of the tonal areas. In the right
hand image, you can see that almost all pixels fall into two categories:
dark area and light area. This is the sign of a true black &
white image. In the left hand image, you see that there's definitely
a peak for dark, and less so for light, there's a lot of "in between" pixels
as well. This is an indicator of a greyscale image.
|
For simple correction of a b&w or greyscale image, you can generally
get good results by moving the low & high sliders to a point within
the peak dark & light pixel range. (The middle slider controls
weighting towards light or dark. If you really need to use this,
it might also be time to consider using the curve tool.)
To the left, the top level tool gives an example of where to move the sliders. (Underneat is the histogram after the level tool has been applied). It is possible to get a similar result simply by cranking up contrast, but you will gain better control over the finished image by using the levels tool. Most greyscale images can also be fixed via the Levels tool, though under certain circumstances, you might be better off with the curve tool. (For example, with a true half-toned image, you might be better off using a stair-step, curve.) Note that with some greyscale images, there may not be any pixels that are supposed to be absolutely black or absolutely white! So be very careful how you do adjustments. (below: another example of the magic of level tool on a b&w image)
|
Normally when working on a b&w image, I do the level correction first, then fix any remaining problems, then resize down. For a greyscale image, I correct image errors first, then correct levels, the resize down.
B&W images (and half-toned images) are 8-bit imags. This means they are stored as 256 (0 through 255) levels of lightness/darkness. Many b&w images were in fact created for printing as a 1-bit image -- absolute black, or absolute white.
This means that you can do some intelligent palette reduction to more
efficiently store the image in a lossless format (e.g. GIF), instead of
using lossy jpeg (which usually ends up degrading the picture quality).
I don't want to have to beat anyone's head in about this, but I know that
there are some people who need convincing over the issue of size considerations:
A restricted palette GIF saves smaller & with less image distortion
than an 8-bit jpeg.
figure 1 b&w lines (7 color) - gif: 2182 bytes This image looks like it's black & white, but in fact, it contains 7 shades. Don't believe me? Check it for yourself. |
figure 2 b&w lines (7 color) saved as jpg: 7,819 bytes at quality 5 (ps5.0) This also looks like it's black & white, but now it in fact has 15 shades due to jpeg artifacting. (Artifacting is visually very minimal here, but that's because these are all just parallel lines). |
figure 3 b&w lines (2 color) - gif: 1,547 bytes This is the image after I converted the palette to absolute black & white. You really can't tell the difference from figures 1 & 2, but it's considerably smaller. |
figure 4 b&w lines reduced in size 50% I converted the image from figure 3 to greyscale, then reduced it in each dimention by 50%. Now it's 7 colors again! Why? because a black pixel next to a white pixel, when reduced, are averaged together -- this keeps the general tone of the image visually more similar to the larger version, and ends up being a LOT better than simply throwing every other pixel away. So be aware that the more you reduce a b&w image in size, the more levels of grey you need to store to maintain the same general appearance as the larger image! |
Here's how to get Photoshop to do a "fit to" palette reduction:
figure 1: 8bit gs as GIF: 66,631 bytes |
Image -> Mode -> RGB Color
Image -> Mode -> Indexed Color (you have to do it in this order to get the palette window...) change "Exact" to "Custom" and you will then get to define the color table to which you will map your image. Note! This is VERY different from the sequence
|
Many
b&w and greyscale images will map into an evenly graded 4 bit (16 shades)
table. So. First step, left click+hold on the upper right corner,
and drag one row over.
(notice that the top row colors are highlighted) You'll then get a color picker for "first color" for which you should set r,g,b values to 255,255,255 (pure white). For "last color" set r,g,b values to 0,0,0 (pure black). Then, select everything from the second row to the end, and set both first & last colors to 0,0,0 (all black). You want a color table that looks like this: |
figure 2: 4 bit gs as GIF: 24,012 bytes This is probably the best compromise between size and quality.
|
figure 3: 3 bit gs as GIF: 16,295 bytes 25% smaller than 4 bit, but the reduced color palette is pretty noticeable in some spots (look at the shading beneath the bubble). |
figure 4: 8 bit gs as JPG (quality 5 ps5.0): 29,634 bytes Just about indistinguishable from figure 1, plus a whole lot smaller. On the other hand, there also isn't a heck of a lot of difference between this and figure 2, and GIF doesn't introduce any artifacting. |
Note that palette shades for indexed colors are stored in terms of the number of bits it takes to represent them.
1 bit = 2 shades, 2 bit = 4 shades, 3 bit = 8 shade, 4 bit = 16 shades, 5 bit = 32 shades, 6 bit = 64 shades, 7 bit = 128 shade, 8 bit = 256 shades
It makes absolutely no sense to restrict yourself to, for example, 20 shades, because GIF storage will still be 5 bits per pixel. If your image requires more than 32 shades to prevent degredation, you might as well be using jpeg compression.
Because the goal is to create a high quality image archive of our comics, image resolution and color quality are important to us. Through a general consensus, we've agreed on the following: