The market's been flooded with good, cheap color scanners, and chances are one will end up on your desk soon if it hasn't already. A scanner opens up a world of possibilities. You can pour scanned images into reports, company newsletters, Web pages, and presentations. The thing will even convert typed pages into word processing documents via OCR (optical character recognition).

But while you can get a great scanner for as little as $400, the so-called automatic scanning software that comes with it isn't particularly smart. Oh sure, the automated features will get you a scan, but they won't always deliver the best scan. Hence the pros who've been there offer this 15-minute Scanning 101 class. Industry gurus like author/designer Daniel Will-Harris, Adobe Systems creative director Russell Brown, photo service bureau owner and photographer Andrew Rodney, authors and publishing whizzes David Blatner and Steven Fraser, and yours truly will show you the ropes.

WHAT'S A GOOD SCAN?

Most beginners think a good scan is one that tingles your retinas with blinding contrast and vine-ripened, saturated colors. Professionals, however, know the reverse is probably true. A somewhat flat-looking scan is usually better, because it will probably contain the maximum range of raw digital color data your software, printer, or display ultimately needs to produce the truest rendition of the original image. The scanning software's automatic contrast and color settings that produced those punchy, high-contrast scans probably kissed off important image detail in the shadows or bright areas (highlights), details you might need when adjusting the image later.

THE ORDER OF THE SCAN

While not a secret society, the Order of the Scan is the righteous path to quality scanning. You have to know what to do and when, in the right order. As you'll see, getting the best scan involves making as many adjustments before the final scan as you can and performing select tasks in an image editing program (such as Photoshop, which we recommend you get) after the fact.

Here's our secret order of good scanning in summary form, which our faculty will tackle in greater detail in the step-by-step, hands-on section that follows. Steps one through 10 take place before the final scan; steps 11 through 16 come after. Start with sharp, clear, properly exposed original material. The best scanner and software can't work miracles. Know the destination of your scan. Is it headed for a color laser printer? Photo-quality service bureau output? An on-screen presentation or Web page? Several of the above? Where the scanned image ends up will dictate how you scan it. Know the nature of your subject. Is it a color photo, black-and-white photo, black-and-white line art, or something in between, like a charcoal drawing? It will make a difference in how you choose to scan. Based on the previous information, calculate the resolution you'll need to properly produce the desired output. Depending on the situation, you may use this resolution when you scan, or you might scan at the unit's maximum optical resolution and change the resolution later in an image editor. Place your material squarely in the scanner or with the same rotation you want in the final result. Select the correct scanning mode, be it line art (black-and-white), grayscale (256 levels of gray), or RGB color (color photo quality). Take a preview scan and select only the portion of the image you want in the final scan. This will save time and keep the file size down. If your scanner software allows it, adjust the contrast and brightness using the gamma curve setting. Avoid brightness/contrast sliders. If you can, set the white point and black point, which will be described later. Adjust the separate red, green, or blue color channels to get the right color balance and saturation. Now make the final scan and save the unedited result in either your image editor's native format, such as PSD (Photoshop) or TIFF (Tagged Image File Format). Make additional adjustments to gamma, white point/black point, and color, in that order, using your image editor. If necessary, adjust the image's resolution. This process is sometimes referred to as downsizing or downsampling. Apply sharpening. If your editing program has it, use the unsharp mask feature (which, contrary to its name, lets you precisely sharpen an image). If a small file size is important because the image will be used on a Web page, reduce the number of colors from, say, 24-bit RGB (16 million colors) to 8-bit (256 colors). Save the final image in TIFF format for archive purposes, then save it in the smaller, compressed JPEG format for distribution.

GOOD BEGINNINGS:
STEPS ONE THROUGH THREE

With the previous overview in mind, you're ready to get the nitty gritty on what to do, how to do it, and when.

Getting a sharp original to scan, determining the scan's ultimate destination, and figuring out what kind of original you're scanning isn't complicated, but it's crucial to the end result. The highest resolution scanner in the world won't make a blurry, poorly focused color photo much sharper. If the print is also overexposed, color details in the light areas of the photo will be scanned as white because there's almost no information for the scanner to record. Conversely, details in an underexposed print may turn black.

How to scan the original is also more complicated these days, since the same image may be used in a printed company newsletter, on a corporate Web page, and in a business presentation. It's important to know ahead of time where an image will end up, since each destination requires different scanner settings. Resolution and color depth differ at the very least. But you don't want to scan the same original three times. It's better to scan once and end up with a file you can successfully and efficiently modify to fill all three purposes.

Scanning tips. Illustrations that are strictly black-and-white (such as etchings and engravings done in black ink) and text pages headed for an OCR program should be scanned in line art or black-and-white mode. The scanner will simply record whether any given pixel is black or white. There will be no grays in between. For black-and-white photos, pencil sketches, charcoal drawings, ink washes, and aguatints, the picture is usually a series of gray tones. Scan these using the grayscale option. The scanner will record 256 levels of gray (8 bits of data) for each pixel, including those that are black or white. For color photographs, where a huge number of colors blend, scan using the true-color, RGB Color, or 24-bit color setting. For each pixel, the scanner records one of more than 16 million different colors--24 bits of data per pixel. As you might expect, the result is some pretty large image files. For example, a 150dpi scan of a three-by-four-inch snapshot can easily consume 790K of disk space. The same scan of an eight-inch-by-10-inch color shot could take up to 5MB. Some color illustrations have a relatively small number of colors: several dozen or less. To minimize file size, you may want to scan these originals as 256-color files. This scanning mode codes one of 256 colors (8 bits of data) for each pixel. This mode is often called indexed color because these 256 colors are chosen from a palette of 16 million colors. The GIF images commonly used on Web pages are stored in this format. Professional tip

Beware of 256-color mode. You'll get better results--especially with color photos--if you scan the image in 24-bit, RGB color and then convert it to 256 colors using Photoshop or a similar image editor. (See the discussion below.)

--Bob Weibel

CALCULATE THE IMAGE
RESOLUTION: STEP FOUR

Once you know what kind of original you're scanning and where it's headed, it's time to figure out the appropriate scanning resolution. Remember resolution refers to the details your scanner can record. Scanners turn your photos, drawings, or text into rows of tiny, tightly packed digital dots, which are stored as bits in a file. Like a mosaic of tiles, these dots, each a single color, form the image. Files of these dot arrays are called bitmapped or raster files. The tighter the dots are packed, the greater the image resolution (in dots or pixels per inch), and the sharper the scanned image.

Scanning tips. Line art looks best when scanned at the resolution of your output device. If you're printing a black-and-white illustration on a 600dpi LaserJet 5, scan the image at 600dpi. If you scanner's optical resolution tops out at 300dpi, use the scanning software's interpolation feature to boost the scan to 600dpi. If you have a top notch image editor like Photoshop, use its interpolation feature instead. Note: Your laser printer's resolution enhancement technology might provide enough smoothing to make even a 300dpi scan look sharp. Do some test printouts and see. Dye-sublimation and Fuji Pictrography printers found in service bureaus can print color as a continuous tone. That means they can lay down one of 16 million colors at each printer spot. The output is positively photographic, without the halftone patterns you get with most inkjet and color laser printers. If you're scanning pictures that contain text, fine textures, 3-D renderings with fine lines, or fabrics, scan at the printer's resolution (in this case, 200dpi or 300dpi) for the best results. Otherwise, you can probably do just fine with a 200dpi scan. Printing presses and most desktop color and monochrome printers use clusters of ink dots--either halftones or dither patterns--to create the illusion of continuous color or grayscales. Halftones are repeating patterns. Their levels of fineness are measured in lines per inch, often referred to as line screen frequency. The higher the number, the finer the image. Dither patterns use a more random pattern of dots. All of our experts agree that regardless of the final output device's resolution, you need only to scan a color or grayscale image at 1.5 to 2 times that of the device's line screen frequency. Thus, if your 600dpi monochrome printer uses an 80 line-per-inch screening, a 120dpi scan should be fine. (For more information, see the "Recommended Scanning Resolutions" table on page XX.) Professional tip

Beware of interpolated resolution. Scanner software often lets you scan above the unit's true optical resolution using an algorithm that estimates the placement and color of the pixels it missed. But interpolated scans, especially with scanner software, can look fuzzy. If you must boost a scan's resolution, use the routines built into Photoshop or a similar program.

--Russell Brown

Professional tip

The same is true if you want to lower the resolution of an image for display on the Web or for output on a low-resolution printer. Scan the image at the scanner's maximum optical resolution and then downsample the result to a lower resolution using Photoshop or similar image editor. Why? For starters, most scanners are optimized to do the best job at their highest resolutions. Just as important, good image editing programs do a better job of estimating the correct color of interpolated pixels.

--Russell Brown and Bob Weibel

GET IN POSITION: STEP FIVE

Don't just plunk your original down on the scanner's flatbed and hope for the best. Line up the original carefully so you don't end up with a tilted scan. Either align one side of the original against an edge of the bed or use a drafting triangle to get it square. You don't want to waste time straightening a crooked scan in software.

Professional tip

Pre-rotate. If you know you're going to rotate a scanned image, save time and rotate the original on the flatbed, using a protractor as a guide. Rotating via software is not only time consuming, it can degrade the image quality.

--Bob Weibel

SNEAK PREVIEW:
STEPS SIX AND SEVEN

Most scanning programs can take a quick, low-resolution scan of the entire scanning bed and then let you select just those areas you want to capture in the final, high-resolution scan. You can also adjust image brightness, contrast, and color values at this point, so the final scan captures exactly what you need. In short, do as much as you can before you perform the final scan to get the image right. You'll end up with more image data to work with. Fiddling after the scan often results in loss of image data and thus, image detail.

Note: Most scanning programs can't automatically detect whether the original is line art, a photo, or a peanut butter sandwich, so make the appropriate selection before running the preview scan.

Another note: Most scanners come with TWAIN drivers or Photoshop plug-ins that you scan directly into your image editor or page layout program. So instead of launching a separate scanning program, you simply select File*Acquire in your target program; select the TWAIN source (namely, the scanner); and the scanning software will pop up, let you make adjustments, and do the job.

BRIGHTNESS AND CONTRAST:
STEP EIGHT

It's obvious when a scan looks too dark, too bright, or too sharp, but fixing the problem isn't so simple. The brightness controls offered by your scanning and image-editing program only control the overall shading. Crank it up and everything gets lighter, which isn't necessarily good.

Contrast deals with the range of differences between the brightest and darkest parts of an image. An image with overall high contrast has dark and light values at the far ends of the scale--namely, black and white. An image with low contrasts has a smaller range of light and dark extremes, but the image may look flat. If you crank the contrast control up, you may bring out details in the shadows, but anything that's already fairly light will be washed out. Turn the contrast down, and details in the shadows turn to mud.

Naturally, if your original photo was badly over- or under-exposed, there's only so much your scanner can do. But even in the best of circumstances, most everyday scanners have a hard time capturing detail in the lightest and darkest areas of an image.

Professional tip

Don't slide. Avoid the brightness and contrast slider bars in scanner and image-editing software. They make linear adjustments which, unfortunately, don't jibe with the way brightness and contrast naturally shift in an image. Instead, use the gamma curve feature in your image-editing program, which changes these values in a nonlinear way. For example, it shifts tonal values more in the midrange than at the light and dark extremes. In Photoshop, the gamma curve function is found at Image*Adjust*Curves. Most image editors let you create a custom curve--a specific contrast/brightness setting unique to a particular image--and save it so you can use it again.

Gamma curves work surprisingly well for most scanned grayscale or color images. To apply a gamma curve, simply indicate a gamma value (commonly between one and two). The scanner software or image editor automatically applies it. Most photos look good between 1.2 and 1.6.

--Andrew Rodney

WHITE POINT/BLACK POINT:
STEP NINE

Every image you scan, especially grayscale and RGB color images, has an optimal range of useful tonal values. But often parts of an image are too dark or light for a monitor or printer to render as anything other than 100 percent black or white. But what if you could remap the extremes so they fell within a tonal range your output device could produce? That's the idea behind white point and black point settings, which often clear up some color problems as well.

For example, an 8-bit grayscale image has a white to black range of 256 values. (An RGB color image would have red, green, and blue ranges.) Instead of spreading all those values from the blackest black to the whitest white, you can bring in the boundaries a bit, setting the low at 40 (a dark gray) and the high at 238 (a light gray). Every gray in the image below 40 is considered black; every gray above 238, white. This way, you don't waste the scanner's precious 256 tonal values on areas where there's no important detail to reveal. By packing those 256 values into a narrower visual range, you get better differentiation among those grays. Instead of your image looking flat, the details pop out much better.

Depending on your scanner software, you typically set the white and black points using either an eye dropper or a histogram. For example, with the VistaScan software that comes with Umax scanners, you get white and black eyedropper tools. Click the white one and drag it over your preview scan. You'll see rapid changes in the highlight gray number (the tonal level reading) in the lower left corner. Cruise around with the dropper until you find the highest value that still shows some visible detail. Since the highest possible value is 255, expect this number to be a bit lower. Then click to set the white point. Select the black-point eyedropper and repeat the procedure, finding the darkest area of the image that still reveals some detail. Try several selections until you find values you like.

Histograms provide a map of the range of grayscale values in an image. The ruler along the bottom shows the entire 0 to 255 (black to white) tonal range. (An RGB color image has three ranges: one each for the red, green, and blue values. But most scanning software also displays a single combination histogram.)

The mountains and valleys in the histogram are really a graph that shows how many pixels are assigned to a particular gray value. The histogram above hasn't been adjusted, so you can see there are white and dark areas in the image far outside where most of the gray values reside. To change the white and black points, click on the white or black triangles and drag them toward the point where the graph starts to climb, as shown in this screen.

The aim of setting the white and black points is to focus the scanner's tonal range on the most important part of the image. This technique forces the scanner to ferret out the details in this range. The result is an image with better contrast.

Professional tip

Use the pixel-poppin' dropper. If you can, zoom in on the preview scan when setting the white or black point with the eyedropper tool. You'll have a better chance of hitting the value that will serve you best at each extreme.

--Andrew Rodney

COLOR ADJUSTMENT: STEP 10

It's fairly common for scanned color photos to end up with unwanted color tinges known as color casts. The original photo may have a cast that becomes more noticeable when you scan it, or your scanner/software combination may have introduced it. For example, the healthy glow of a scanned flesh tone may look fine until you check the original and see that the scan is really too pink.

Entire books have been written about color correction, but we can at least help you get your colors reasonably in tune. Professionals use monitors, scanners, and printers that are carefully calibrated to provide good color matching, but most of us have to fudge a bit. One simple way is to make some sample prints to get an idea how your printer's grayscale or color output differs from you monitor's. Top notch image editors like Photoshop have calibration routines that will help you get a decent color match with a printed sample.

Professional tip

Remember the white-point/black-point fix. Many color casts disappear when you assign the black point and white point as described above. That's why you should do tonal tweaks first and then fiddle with color.

--Andrew Rodney

Professional tip

Histogram channels are another option. If the white and black point setting doesn't eliminate a color cast, switch to histogram controls. When viewing color histograms in your scanning software or image editor, always adjust the individual channels (red, green, blue) first, not the master RGB control. Select a color channel, and then move the arrows for the black point and the white point so they rest where the mountain range in the histogram begins to climb. Do this for all three colors and then check the results. This trick usually does the job.

--Russell Brown

FINAL SCAN: STEPS 11 AND 12

When your preview scan looks about right, click the Scan button for the final scan. Chances are you'll be scanning from within an image-editing program like Photoshop, PhotoDeluxe, or Paint Shop Pro, so the completed scan will automatically appear in the application. Print out the scan. If you have to fiddle with the image's tonal range and color to get the right output, start over. Apply those changes with the scanner's software and redo the scan. If you manipulate a scan too much after the fact, you can lose precious image data, and that means an uglier final product.

I like to save that raw scan as a backup and then use a copy for the final tweaks described below. Save the scan in the image editor's native format or in TIFF, using LZW compression.

Professional tip

Don't pig out on JPEG. While JPEG compression can easily squash an image file to one-tenth its size without visible degradation, it does lose image data. This can make a difference if you later tweak the file in an image editor. So only use JPEG to distribute final versions of the image.

--Daniel Will-Harris & Andrew Rodney

DOWNSIZING: STEP 13

If you want the scanned image to ultimately appear smaller onscreen or on the printed page, now's the time to shrink it. Shrinking the image's dimensions will naturally shrink its file size, too. For example, at 96dpi (typical PC screen resolution), a two-inch-square image is 192 pixels by 192 pixels. Shrink the image to one inch square and you shave the size to 96 pixels by 96 pixels. Don't worry about dpi; this information is ignored in the GIF files typically used for Web pages. Think of the image in terms of pixel dimensions.

To downsize an image--in short, to remove pixels--in Photoshop, select Image*Image Size. To help you think in terms of pixel dimensions when resizing images, work with the Pixel dimension portion of the Image size dialog box and ignore the Print size section.

If you're certain you'll only need a lower-resolution image, do this downsizing between steps seven and eight. You might as well work with a low-resolution image since it will redraw much faster.

Professional tip

Online '96. When downsizing images for online and presentation display, select 96dpi for the Print size resolution (the Windows standard) instead of 72dpi (the Mac standard). There are far more PC users surfing the Web, and Mac users won't notice any difference.

--Daniel Will-Harris

SHARPENING: STEP 14

Without going into technicalities, the scanning process tends to soften the edges of images, so you need a tool to resharpen them. Scanning at a higher resolution won't solve the problem. Why shouldn't you sharpen earlier? Because the technique works on the pixels present in the image; it doesn't make sense to sharpen an image before you remove or add a bunch of pixels. I also prefer to sharpen after I've made the final tonal and color adjustments, since these affect the visible edges in an image.

Our best advice is don't overdo it! If you apply a Sharpen More filter and notice slight halos around objects or puckered textures that were once smooth, hit Undo. To best judge the effects of sharpening, view the image at 100 percent (a 1:1 ratio).

Professional tip

Use Unsharp Mask (USM). Ignore the misleading name. This sharpening filter found in Photoshop and similar programs lets you control the degree of sharpening. Functions like Sharpen, Sharpen More, or Sharpen Edges don't. There are three settings to consider in most USM filters: Amount, Radius, and Threshold.

Amount indicates the intensity of the USM's edge enhancement. Start out with values between 200 and 400 (at least in Photoshop) and work down from there. Radius sets the width of the edge effect. The wider the radius, the more obvious the sharpening. Depending on the image and its size, try Radius values between .6 and 2.4. The Threshold setting specifies how far apart two pixels' tonal values must be before the filter affects them. A smaller value results in more sharpening. Start with a low value and work up from there.

--David Blatner and Steven Fraser

Professional tip

Sharpen before JPEG. Apply sharpening before you save images in JPEG format. If you sharpen after, little quirks introduced by JPEG compression appear, such as small tiles in the image.

--Bob Weibel and Andrew Rodney

COLOR REDUCTION: STEP 15

Since large, 24-bit color files are often unmanageable for Web pages and multimedia presentations, people convert them to 8-bit (256-color), indexed color files, as described earlier. You naturally lose image data, but to get the best image possible, use your image editor's adaptive palette function.

This automatic feature surveys your image to determine which colors are being used and creates a custom palette of the closest matching 256 colors from a 16-million-color master palette. You can help hone the palette by selecting an area of the image that contains the colors you want.

If the scanned image only has a few colors, set the color depth to fewer than 8 bits and save disk space and download time. A 4-bit color depth will code for 16 colors; a 6-bit color depth will code for 64 colors.

ARCHIVE IT!: STEP 16

Don't go to all the previous trouble, especially with grayscale and color scans, and then save the only copy in JPEG format. Save it in a standard format--we recommend TIFF--that doesn't lose image data. If you're worried about soaking up disk space, use your editor's LZW compression, which often reduces file size by 50 percent or more.

POST GRAD WORK

You now have the basic tools to start buffing up your scans and confidently exploring new areas. As in any course, however, there's some required reading. One of the best overall guides to scanning and image handling is "Real World Photoshop" by David Blatner and Bruce Fraser. It is edited by Stephen Roth and published by Peachpit Press. For an in-depth look at using histograms and other image-editing tools, check out Barry Haynes' "Photoshop 4, A Master Class for Photographers, Artists, and Production Artists" published by New Riders. May you scan well and prosper.

© 1997 Bob Weibel. All rights reserved.
Bob Weibel is a writer in Ashland, Ore., a former senior technical editor for Publish, and the coauthor of "Desktop Publishing Secrets" and "The QuarkXPress Book, Windows Edition" (both from Peachpit Press). He also writes Computer Currents' Publishing Advisor column. You can visit Andrew Rodney at www.camera-darkroom.com. Check out Daniel Will-Harris' Web site at www.will-harris.com.

Txt Intre Wurds: Getting the Most Out of OCR

Scanning is a small but important part of optical character recognition (OCR). The bulk of your efforts will involve using the OCR program's options--distinguishing text from graphics and teaching the program to better recognize a typeface--after you've scanned the page. But some basic scanning strategies do apply.

Even if you have a 600dpi scanner, you should probably scan text pages at 300dpi or 400dpi resolution for OCR, using the line art (black-and-white) mode. But the end result largely depends on the document you're scanning. I've gotten good OCR results from 200dpi text scans, but these were simple, single-column pages printed in crisp 12-point Courier.

Your OCR scanning technique pays off most when you're trying to capture poorly printed pages. Look closely at that photocopy of a photocopy and you'll see faded or broken letters. These will really confound the OCR program, which may either miss the faded character altogether or try to interpret a broken letter as two separate characters. For example, if there's a horizontal break in all the lowercase Ts on a page, the OCR program may interpret them as lowercase Ls or 1s.

Adjusting the scanning software's brightness and contrast controls before the scan can help. Take a preview scan and zoom in to examine the scanned characters in detail. You'll immediately see where portions of the characters have faded to white, resulting in breaks. If you see broken characters, turn the brightness setting down 10 percent and do another preview scan to see if the increased darkness has filled in the broken outlines. If not, lower the brightness another 10 percent. When the breaks finally fill in, see if the increased darkness has also created random blotches on the page. If so, crank up the contrast to reduce the blemishes. Save these settings and take the final scan of this and the other pages.

In extreme cases, preview scan the page as grayscale and manipulate darkness and contrast settings until the characters stand out adequately. Then take the final scan and convert it to black-and-white in an image editor. In most cases, though, this process is hardly worth the time and effort, since you could probably type in the page more easily and accurately. And therein lies the problem with OCR. Unless you're going to convert many similarly printed pages, the time it takes to scan the original, fine-tune your OCR program, and proofread the final output may simply not be worth the effort.

--BW

Recommended Scanning Resolutions

Here are recommended scanning resolutions for several common printers and printing services.

OutputOptimal scan resolution (dots per inch) Line artGrayscale/Continuous tone color LaserJet 5M (600dpi, laser) 600dpi150dpi-200dpi Epson Stylus Color (720dpi, inkjet)720dpi 180dpi-240dpi HP DeskWriter 680C
(600dpi B&W, 300dpi color, inkjet) 600dpi 100dpi-150dpi Color LaserJet 5M (300dpi) 300dpi100dpi-150dpi Fuji & Dye-sublimation printers (200dpi-300dpi)n/a 200dpi for images; 300dpi if text is present. Web page (72dpi-96dpi) 72dpi or 96dpi 72dpi or 96dpi