Not all digital images are equal. Some are so small that they are adequate for web purposes only, while others are suitable for the more rarified heights of print. There are a few simple rules of thumb to be employed when judging which is which and how much an image can be enlarged without ruining it.

Image Building Blocks

The ancient Greeks were the first to guess that if you cut matter into small enough pieces, you would eventually end up with the fundamental particles from which all things are composed. Digital images are similar. Peer closely enough, and every photograph resolves into a grid of tiny dots. On a computer screen, each point of colour (or pixel) is composed of three tiny image elements. The three elements are red, green and blue, and when each is illuminated in various combinations at one of 255 levels, give rise to one of many millions of potential colours. Back out at the human scale, viewers see a seamless blend of colour, detail and motion.

Dots Per Inch

The pixel density on a typical computer screen is 72 pixels per inch, which means 72 pixels on each side of a square inch, yielding a total of 5184 pixels per square inch. Images for use on the Internet are hence set up at 72 pixels per inch. Unless the absolute pixel size of an image is quite large, it is relatively unusual for a web-optimised image to be suitable for print purposes. Print images usually require a minimum of 300 dots per square inch, which equals 90,000 pixels per inch squared. Hence, print demands much larger file sizes.

Interpolation

All digital photos are comprised of a grid of pixels, also known as a raster image. When a digital image is enlarged, the image software automatically adds (interpolates) additional pixels. If enlarged too much, the resultant image can become noticeably blurry. Judicious use of image sharpening software can correct this to an extent, but not if there wasn't enough information to begin with. Also. many jpeg images optimised for web purposes suffer from the effects of 'lossy' image compression, where image information is discarded and the picture becomes 'blocky'. Sharpening an image degraded in this way can be counterproductive.

For example:

An 300dpi image that is 500 pixels along each side translates into a 4.2cms on a side print image. An 72dpi image that is 2000 pixels on a side works out as a 17cms on each side image when resized (without resampling) to 300 dpi. Hence the image that is larger in terms of absolute pixel dimensions is always the best option, even if it is nominally at 72dpi.

Resize without Resampling

If using a good quality image editor such as Photoshop, it is possible to resize an image without resampling (interpolating) it. In other words, you are able to change the size at which an image will print without altering the overall number of pixels. The value that changes is dots per inch. For example, a 1000 pixel image at 72 dpi would print out at 35cms across. If the resolution is reset to 200dpi without resampling, the image will now print at 12.7cms across. At 300dpi, the image will be 8.47cms across. The same image, the same overall width in pixels, but a changing relationship to the resolution, and hence the print size.

Example from Photoshop CS3:

Note the pixel size, print size and resolution

The pixel size remains the same, but the relationship between the print size and resolution has changed

Summary:

Try to use images at, or close to, the original size unless there is a lot of image detail to work with.

Look for images with large pixel sizes (1000 pixels and above, preferably). Anything smaller than 400 pixels across is likely to be of little use for print.