News from

The Computer Workshop

Know What You're Doing With Photographs

I want a digital camera but I haven't quite convinced myself to spend that kind of money, and besides, I have a scanner. This is for those camera and scanner owners (and prospective owners) who have been asking questions about photos or showing by the size of their files that they should have been asking.

Photos are stored in what is called bit-mapped file format: a map of which dot is what color much like a tile pattern. Common bit-mapped file extensions include BMP, TIF, JPG, GIF, PSD and PCX. Just like enlarging a tile pattern or a newspaper photo, enlarging a bit-map just makes the dots bigger so that lines often appear jaggy; this is called resolution dependent. The other major file format, vectored, contains a mathematical description of the image rather than collection of dots. Created by "draw" type software and used for things which are printed at many different sizes like architectural drawings and logos, these files are not resolution dependent so they can be made any size without getting jaggy (unless they contain a bit-map as part of the file; EMF, WMF, EPS, DXF and DWG are some vectored file types). Photos are not readily reduced to mathematical description so they will always be resolution dependent: to get the best output you need to know the resolution of the output device (screen, printer etc.) and provide enough dots in the original image from your camera or scanner.

The maximum number of dots you can capture is maximum resolution (dots per inch, dpi) multiplied by the area (sq. inches) of the image. In a traditional photograph resolution this is a function of the silver grain size in the film. With a digital camera the maximum number of dots is set by the number of pixels (dots) in the CCD (Charge Coupled device: the equivalent to film in a regular camera). In a scanner it is the number of optical sensors in the image area. Don't be fooled by "interpolated" or "digital zoom": it doesn't give you any more dots. You can achieve these effects yourself with your photo-editing software so why let the camera or scanner do it? The real resolution is the actual or optical resolution given for your device. This is why 2 megapixel cameras (1600 by 1200 pixels = 1,920,000 maximum total dots) cost so much more than 1 megapixel (1024 by 786 pixels = 789,504 total; see for more camera information).

How do you decide what resolution you need?

A 1 inch square image at 200 dots per inch contains the same number of dots (40,000) as a 2 by 2 inch image (= 4 square inches) at 50 dots per inch (( 4 sq in. x 50) squared = 40,000 total dots). In general, if you double the linear measurement, you cut the resolution to one quarter. Size and resolution are two parts of the same thing: the total number of dots in the image. If you capture an image at the right resolution for one purpose and then need it larger, it can get very fuzzy. Some photo-editing software will let you make it bigger and soften the jaggies but it can't add detail. You really will have a better image if you start over. If you don't know what you want to do with an image, a good compromise is to store the image at the maximum size you might need on CD or other removable media to avoid filling your hard drive. Copy it to your hard drive to modify size and resolution for each use. All of the information about resolution to follow assumes your photo is at the correct size.

For screen use, like on your web page, a resolution of 72 dpi and maximum size of 800x600 pixels (screen dots) fills the screen (700x500 is better so you leave an edge for the browser toolbars). Some screens have more total pixels and the number of dots per inch varies slightly so using the lowest common resolution will work for everyone. If you know your audience is using big screens, using a maximum of 1024x768 pixels is justified (small screen owners will have to scroll).

The bit-depth, the number of shades of color in the file, is automatically reduced if your monitor doesn't support it. Most cameras, scanners and monitors use 24 bit color, that is 2 to the 8th power = 256 shades of each of the 3 phosphor colors, Red, Green and Blue. This means that instead of 1 bit for each dot we are really storing 24 bits for each dot: see the effect on file size in the table below. The different operating systems show colors slightly differently so you may want to look at your image on both Mac and Windows as well.

If you want to print the image, things become more complicated. Your camera, scanner and monitor store transmitted color: all colors can be described as a combination of red, green and blue light (RGB) which together make white. Your printer uses cyan, magenta, yellow and black inks to make colors (there are other color modes which use 6 or more ink colors but they are similar in concept). CMY together make black: like you used to do with finger paint. They throw in black ink to avoid saturating the paper and to make black text easier. Converting between the two color modes changes your colors slightly so you only want to do it once (keep your stored originals in RGB so that you can start fresh if needed). Most software converts using a one-size-fits-all approximation but Photoshop and some other photo-editing programs correct the conversion for the specific printer: once again, you need to know where you are going with the image.

Resolution on a printer is not what it seems. Unlike the screen where each dot color is made of shades of each color (RGB), printers can only print ink or no ink. They make shades by a process called dithering. If your printer had a resolution of 64 dpi it might take groups of 4x4 dots and use 4 dots filled in each 4x4 square to be 25% shaded, 8 dots for 50%, 12 for 75% and so forth.

It would be getting 16 shades (4x4) by cutting the real resolution by one quarter (64/4=16dpi). Thus, instead of 64dpi, the printer could only print 16dpi if you wanted 16 shades of gray. It would throw out all the excess. For your 600dpi printer, you only need 170 dpi, anything over this wastes time and space. There is a long equation which determines the optimal number of shades at a given resolution: I've provided a lookup table for easier use. Line screen (lines per inch, lpi) comes from manually halftone screening photos and is now used to describe dithering for print; double it to get scanning resolution (dpi).

Resolution uncompressed 1x1 inch file size
Use lpi dpi 8 bit gray 24 bit RGB CMYK
Screen, Web (1K byte » 8,000 bits) 72 5.2K 15.5K 20.7K
Analog photocopier 80 160 25K 75K 100K
Some digital copiers, Newsprint 85 170 29K 85K 113K
600 dpi Laser printer 85 170 29K 85K 113K
720 dpi inkjet (some go higher) 100 200 40K 118K 157K
quick print (photo-direct plates) 120 240 57K 169K 225K
magazine-inside 133 266 70K 208K 277K
magazine-cover 150 300 88K 264K 352K
Fine art prints, some ads 200 400 157K 469K 625K

As you can see, files get big fast: compression saves space. When saving from your photo-editing software for printing you should usually choose compressed TIF; it reproduces the image exactly. JPG is the most common compression for screen images but it is "lossy": it removes detail from the file that you wouldn't notice on the screen, such as subtle color shifts. The other web format, GIF, gets its compression by limiting the number of colors in the image to 256 or lower.

I hope this makes the issues involved in sizing and saving photos more clear. E-mail the cute baby photos at 72dpi (no more than 800x600 pixels total) in JPG but keep a high resolution RGB original (compressed TIF unless you started in JPG) burned on a CD. You can then print it for the album and, after the colors fade in 2 years, print it again for as long as you can read the CD. (The ink fade is the other reason I haven't yet bought a digital camera: Epson now has archival quality inks but the printers start at $800 and the ink is pricy too.) Please call if you have further questions, 206-523-0872.

copyright 2001 Karen Seymour

Back to Articles index         Back to The Computer Workshop home page

Trademarked names belong to their respective companies and are used for identification with no infringement of trademark intended.