Graphics Primer

Once you get into scanning, you will have to learn about resolution, color depth, the capabilities of various output devices, and graphic file formats.

Resolution

This refers to the number of dots (called pixels) per unit area, typically expressed in dots per inch (dpi). The higher the resolution, the sharper the detail, but also the larger the image file – which in turn means more disk space required, and slower transfers by modem. There’s also no point in using a resolution higher than what the output device can support; more on that later!

The triangular relationship between resolution, area and size is as follows:

Color depth

This refers to the number of different possible colors for each pixel. Higher color depths make for greater realism, but also increase the file size as there are more bits of information in each pixel. Once again, the output device determines the final result, and the point at which greater color depth no longer makes a difference.

The relationship between resolution, color depth, area and file size is as follows:

The common color depths used on the PC are:

Output devices

When scanning and editing graphics, it’s important to keep the target output device characteristics in mind – especially resolution, else you will have a surprise when you see the final image size as used by the device!

You do get 1200 and 1440 dpi printers, but they are often not used at this resolution.

Note that color matching between the radiative RGB model (as seen on screen) and the subtractive CMY model (ink-jet printers) is an ongoing challenge. Colors act differently when mixed between these models; a mixture of red and green light gives yellow and gets brighter (additive or radiative color), where a mixture of red and green ink gives brown and gets darker (subtractive color).

Note also that it is not enough to drop from color to greyscale when preparing for true monochrome output, as shades of grey are not supported. Instead, you need to create a black-and-white image using halftoning, i.e. different sized black dots in different densities, to imply different shades. True halftoning (where size of dots differs) is difficult where resolution is limited, and an alternative to the grid of dots approach is error diffusion, where the grid effect is avoided.

Graphic file formats

There are proprietary formats, which may be obscure or common, and there are standard formats. When interchanging graphic files, you should use standard formats wherever possible; the only proprietary format you have to remember is the Windows Bitmap (.bmp) as this is used for desktop wallpaper.

Many file formats address the conflicting requirements of small file size (and thus rapid transfer through slow modems and display) vs. fine detail and realistic color by using data compression, which may be lossless or lossy.

Lossless compression is what archivers such as WinZip use; it allows the original data to be recreated perfectly from the compressed form – obviously crucial for program code! Lossy compression trades off some accuracy in order to gain far greater compression, and is fine for final output, but the errors can build up into image distortion patterns if used for "work-in-progress" while editing the image.

The common file formats are:

Windows Bitmap (.bmp) - any color depth

Tagged Image File Format (.tif, .tiff) - any color depth

Graphic Interchange Format (.gif) - no color depths over 8-bit (256 color)

Joint Picture Experts Group (.jpg, .jpeg) - any color depth

There are dialects of GIF, which introduced interleaving and animation. Interleaving changes the way the image is drawn as it is loading, so that the whole image is visible with detail added as additional data is transferred; this is a big win with slow data transfer technologies such as modems. Animation is the ability to switch between multiple different images stored within the same file. Another web-friendly part of the GIF standard is transparency, so that the outside of a square image can merge into whatever color background lies behind it.

Graphic tools

The scanning process starts in the same place, no matter what graphic application you are using; the TWAIN (Tool Without An Interesting Name) driver. Think of this as a printer driver than works backwards! That’s what actually acquires the graphic data from your scanner (or camera, or…) and presents it to the application.

Then there are the applications that you might use to edit or display the image…

IView is quick to load and display images, and lets you use the Space and BackSpace keys to step through the images in a given directory. It has also become quite proficient at global color balancing tasks, and does file format conversions. It can also show .avi and .mpg videos, and allows speed adjustment for these.

LView Pro in its free-trial form is a bit naggy, compared to IView, but is excellent for global color manipulation as well as offering pixel editing and file format conversions.

Paint Shop Pro is good for chasing pixels, i.e. touching up blemishes, filtering for sharpness or smoothness, and localized color treatments. The program can show a histogram of color values, and convert file formats.

Tactics

In general, scan in with more than enough detail; choose reasonable resolution and use full color even if final output is to be monochrome, so that you have more control during the editing process. Then, immediately save the file in a file format that does not use lossy compression, i.e. Bitmap (.bmp) or TIFF (.tif), remembering to select this type in the SaveAs dialog (just naming the file as .tif etc. isn’t enough).

Having done that, you can proceed through the editing stages. Frame the image you want and choose Crop to get rid of any rough edges, and then save again.

Next, assess the quality of your scan. This has little to do with what your eye tells you ("looks too dark" etc.); rather consult the color histogram. That shows you the range of data in the image, and should stretch all the way from pure black to pure white. A common mistake is to tweak color/brightness/contrast at the TWAIN level, and lose data; the histogram may show a flat line of no content at one end of the graph. Remember that detail cannot be re-invented once lost!

Now you can do whatever touching up is required at a pixel level, e.g. to paint over paper tears, creases or spots in the original. Your most useful tool here will be the "clone" brush, that preserves pattern trends (few photo images have truly "flat" areas of all the same color). If you do want to splat large areas of the same color, use the color picker ("eye dropper") to select the color off the image and then draw filled rectangles of that color. I find that faster than painting, and safer than filling.

If you need to force a scan back to a few primary colors, e.g. after scanning a diagram rather than a photo, a useful trick is to use the color replacement brush with the same value for "find" and "replace". This tool may work within an adjustable tolerance, i.e. any color close to the "find" is "replaced". By using the same color for "find" and "replace", you effectively force all nearby shades to the same value.

Having done all the above, saving as you go, you can apply focal and global manipulations such as sharpen, blur, or color treatments. Brightness and contrast are the traditional adjustments for "too light", "too dark" or "I can’t see anything", but you may find gamma adjustment preserves detail better (consult the histogram).

Then you may want to adjust color in various ways, either through color tools, or by applying gamma etc. to a particular RGB color channel. For really fiddly color stuff, I’d recommend LView as the tool of choice.

Next, bump down the image size to that required by the target output device by using the Resize or Resample functions, along with Crop to get the shape right. Then you can save the work in the final output file format; usually JPEG or GIF, unless you are creating a .bmp for Windows wallpaper. If sending the file top a repro house, rather use a lossless format such as TIFF (check for dialect compatibility).

You may well find that a lossy format looks better and creates a smaller file than a lossless format with lower resolution. Experiment with the compression level, and view at a high magnification (e.g. x 8) to watch the side-effects. "JPEGism" causes the image to appear constructed of diagonally-shaded tiles, especially when magnified, but deliberately inducing this by selecting massive compression gives insight as to how JPEG compression works!

When printing, you will usually have a choice between stretch-to-fit-page, or as-is. The latter may take several pages if you scanned at too high a resolution (you generally go above 600 dpi only to magnify small areas to be scanned) or show a tiny postage stamp lost in the page if your resolution was too low. Stretch-to-fit will typically have a checkbox to preserve aspect ratio; when selected, the contents of the picture will appear the correct shape but may not fill the page.

Scanning can chew prodigious amounts of disk space, and is a good argument for large hard drives and eventually a CDR writer for archiving or transport. If you need to send graphics to a repro house, you need to preserve as much detail as possible (especially resolution) and may have to explore a bulk storage device that uses disks or tapes that the company can read. Finally, don’t attempt color editing or balance correction in 256 color mode; preferably use TrueColor for this!

 

(C) Chris Quirke, all rights reserved

Back to index