Scanning is a very complex process if you want to produce the
best possible images. Most of the work in scanning comes in post-processing,
not in the actual scanning. I cannot begin to explain everything
there is to know about scanning (I'd have to learn everything
myself, first), but here are some tips that may be useful to you.
Not all these tips will work for you, but hopefully you will find
some that you hadn't thought of.
Of course, I am always looking for more tips myself, so feel
free to e-mail me with ideas. :-) If you find some of these tips
helped you, please e-mail me to let me know. I'd like to know
that my time and effort was not wasted. In addition to these tips,
you will find some helpful information at Shyboy's
website. Look "behind the scenes" to find his scanning
tips.
Have you ever noticed scans where you can see the lettering or other things from the "back side" of the page? The way to avoid this is to use a sheet of black paper as a backing. Place this paper directly behind the page you are scanning, and the "bleed through" will not show. Note that you *may* need to "equalize" the scan afterwards as the black backing may alter the image a bit.
Most PCs and Macs have certain fixed screen sizes (640x480, 800x600, 1024x768, etc.). It doesn't make sense to post images that are larger than these sizes because viewers will then either have to resize them "on the fly," or the image will have to be scrolled to be viewed. Unfortunately, most "on-the-fly" viewer resizing does not use very good algorithms. It is better to post an image that is sized for viewing. In my case, I now use a 1024x768 screen size as my guide. Since most of my images are tall, I resample them to 768 dots high.
Note the difference between "resizing" and "resampling." Resizing usually implies simply throwing away (or adding, if enlarging) occasional pixels to create the new image. Generic programs such as LView, VuePrint, ACDSee, etc. tend to use this technique since it is faster than resampling. Unfortunately, this can often create poor results.
"Resampling," however, usually implies the use of complex mathematical algorithms to "sample" sections of the image and rebuild the image so it still looks good at the new size. All the major commercial software (Photoshop 3.05, Picture Publisher, CorelPaint, etc.) use this technique. I believe that the shareware PaintShop Pro also resamples. If done properly, a "resampled" image will generally be indistinguishable from an original scan of print material.
When you scan original images from print material, you may find that the scanned output is grainy or has lines across it. A method that works well is to scan at a larger size first, do some processing on the larger image, then resample to the final posting/viewing size:
This technique works with many magazine and newspaper images. If your images are not turning out as sharp as you'd like, try scanning at a high dot pitch first. Make your original scan 1,000 or more dots high (I usually scan to produce a 3,000 dot high image). Apply a slight blur effect on the image, then resample the image to the finished size.
For instance, if you have a 5 inch high original, you can set your scanner to 300 dpi. This will produce a 1,500 dot high scan. Then apply a slight blur (usually in the image effects menu of most programs). Finally, resample the image to 600 or 768 dots. The shrinking may appear to "resharpen" the image, as long as you didn't blur the image too much to begin with.
Some people will find that they don't need to blur at all. The main advantage to the blurring is that it will help eliminate scanning "patterns" (cross-hatching, diagonal lines, etc.) in the final image. However, some scanning software that comes with scanners (such as Microtek's Scanwiz) include a "descreen" function that often does a similar job of removing patterns in the scans. It should be noted that such descreening functions tend to blur the image a bit.
Unsharp Mask: After blurring (or after using the descreen function) then resampling to the viewing size, it may be helpful to perform one additional step. Some commercial graphics software have a filter known as an "unsharp mask" or just "unsharpening." This is a confusing and non-intuitive label. In fact, the "unsharp mask" sharpens the image in a very pleasing and helpful manner. Unlike the "sharpen" function, the unsharp filter will perform a complex analysis of the image to make the image appear to be more in focus.
This tip requires lots of experimentation to obtain good results. I don't have a set formula that I can use for my own scans, so I suspect that you will need to try different settings to optimize your output. Note that different source material demands different approaches. What works for a particular magazine image may not work with a different magazine or with a photograph.
If you are scanning from print material (newspapers, magazines, books, catalogs), you may see patterns in your scans. These "inteference" patterns (Moire' patterns) are a result of the print material being printed at a certain DPI, and your scanner scanning at an incompatible DPI. Sometimes all you need to do alter the scanning DPI. Of course, if your scanning software has a descreen function, that will usually work best.
Huh? Actually, what this more-or-less means is to make "black" appear black, and "white" appear white, and everything else appear correctly. If you see an image and it looks "washed out," it often needs this "fix." Here's some general background...
Each pixel color (red, green, blue) is represented at one of 256 levels (0-255). That is, red can be at intensity 0 (no red at all in that pixel), 255 (full blast), or something in between. Similarly, green and blue can also be represented at 256 levels each. Since the three colors combine to make up each pixel, you can have 256 x 256 x 256 = 16,777,216 color possibilities. Since each color requires 8-bits to get 256 variations, three colors require three-times-eight, or 24-bits to get the 16M colors.
Okay, so what does that mean here. Well, sometimes when scanning pictures from print material, the scanned image does not utilize all 256 levels. This can be because the "white" paper was not white, or the "black" ink was not deep enough, etc. So what "equalizing" will do is "stretch" the image so that the levels are fully utilizing 0-255 levels.
Luckily, most of the major programs (Photoshop, Picture Publisher, Corel Photopaint, Paint Shop Pro) have *automatic* functions for doing this correction. Ninety percent of the time, this automatic function makes the correct decision, and the image is corrected.
Unfortunately, all the programs bury this function someplace. Here are the locations of the function in some of the major Windows programs:
Adobe PhotoShop - image/adjust/levels <auto> button
After hitting <auto>, adjust manually using the slides Picture Publisher - map/tonal balance/auto clip
Before using the <auto clip> button, go into options and change the default loss from 5% each to 0%. After hitting button, use the slides to fine tune.Corel PhotoPaint 5 - effects/tone/equalize
Paint Shop Pro (shareware) - colors/histogram functions/stretch
If you are using another program, search the help file for the words "equalize" or "histogram." Sometimes it's in "tonal correction." My experience shows that Photoshop's <auto-level> does the best job of this.
If an image seems too bright or too dark, the first impulse of beginners is to look for the "brightness" setting. This is understandable, since we were all raised with televisions. This is a mistake. There is a preferred adjustment known as "gamma." Let me explain...
Above I explained that there are 256 levels of intensity, from 0 to 255. Let's assume that you've "equalized" the image so the range of levels spans the entire 0-255. Here's how "brightness" and "gamma" affect this...
If you increase the brightness level by 15, *every* pixel will be increased by 15. Thus, something at level 0 will become level 15, something at level 150 becomes level 165, and something at level 240 becomes level 255. Thus the entire image is certainly "brighter-looking." So what's the problem? Well, 240 becomes 255, but so does 245, 250, and 255, since 255 is the maximum. Thus all the brightest pixels "blend together" as one brightness level. Also, since 0 becomes 15, the entire range is compressed to 15-255, and the images begins to look "washed-out" again, as there is no longer any true "black."
Gamma, on the other hand, "anchors" the ends (0 stays at 0, 255 stays at 255), while the middle levels are moved proportionately toward the direction you choose. Thus, if you increase the gamma, 0 stays at 0, 255 stays at 255, 126 may move to 140, and 25 might move to 40, but 245 may move toward 250, but never to 255 itself. Thus the entire *range* of levels is intact, but skewed to one side.
This adjustment control is definitely worth trying. *All* image editors have a gamma adjustment setting - search the help file for "gamma."
A clone tool makes the best tool for touch-up. It would be too hard to explain here, but look up "clone" in your help file. The major programs all have this tool, but implement it differently. More than any other tool, this one takes practice, practice, and more practice. Shyboy's page discusses cloning in some detail.
The "normal" file format used for posting to ABPC is the JPEG format (see Q&A below). Since JPEG is a "lossy" format, you get to decide what loss setting to give the file. If you are planning on saving/reloading/resaving/... use another 24-bit format first (such as BMP, TIF, TGA, PCX). Only when you are ready to prepare the file for posting should you convert the file to JPEG. I use TIF as my "work" format. I believe Mary Jane uses TGA.
The reason for not *posting* TIF, TGA, PNG, or other non-loss format has to do with file size. A "typical" TIF file might be 800K to 1M in size, while in JPEG format, the same file might only be 80- 100K in size.
Since you can choose the amount of loss to use in the JPEG, you must decide to balance the smaller file size with the loss of detail in the image. Many inexperienced posters save with too high a setting (too much loss), and the negative results are noticeable. Others save with little or no loss settings, which in practice affords little to no advantage, visually, at the cost of a larger file size.
As with lots of things, experiment to see what works for you. Most programs will save JPEGs at a recommended setting. I tend to take this setting and save at a slightly higher quality. So if they suggest a setting of 75 (out of 100), I might use 80 or 85 for posting.
Note that some programs have some odd quirks. For instance, if I load a JPEG file (previously saved at a moderate loss setting) into PhotoStyler then try to save the file back to disk, the program will automatically save it at a high-loss setting, without warning. That's another reason to "work" in TIF (or TGA or ...).
 |
 |
 |
 |
 |
