In Understanding JPGs: deterioration with multiple saves, I show the ways JPGs can lose quality by the mere process of saving. It also helps to understand a little about how JPGs work.

Think of pixels as tiny squares

When you scan, your scanner records color hue and color density information in an array of samples. I find it easier to think of the samples as tiny squares. By changing your resolution, you change how much area gets averaged into each sample.

Purists will point out that scanners do not actually sample pixels. Instead, monitors interpret information from the sampled areas and display them as pixels. I do not care whether you call them samples, pixels or dots. Like most other people, I use the terms interchangeably.

This example is a detail of a tiny portion of a vignette, scanned at 150 spi (samples per inch) and then greatly enlarged. Your monitor displays the samples as series of squares representing pixels. Each pixel contains information about color. Note that lines that would be sharp in the original engraving are rendered in the scan as jagged lines or collections of squares of different shades.

The example above was enlarged from the image below. The detail originally contained 3,657 pixels, while the one below contains 54,720 pixels.

JPGs save space by compressing images.

The nice thing about JPGs is that almost all scanners and image manipulation software allow you to vary the amount of compression to match your needs. As you increase the amount of compression, the quality of your images and your files sizes decrease. Here are examples of letters greatly enlarged from a bond. (The nomenclature shown below each images comes from my Epson scanning software; other programs use different terms.)

0 compression (100% quality)	20 compression (80% quality)	40 compression (60% quality)
60 compression (40% quality)	80 compression (20% quality)	100 compression (0% quality)

Notice the sparkly halos around the edges of letters. The effect is called 'JPG artifact' and grows more severe as compression increases. The effect is very noticeable here and probably makes you say, "Yuk!"

But, let's pull back a bit. High compression is not always distracting with larger images. If you are after small file sizes for use on the web, don't be afraid to push compression into the 70 to 80 range. Incidentally, there is a w-i-d-e range of nomenclature used by the various programs to express compression. Different programs use different scales and verbal descriptions to quantify compression. The most popular numerical scales are 1 to 10, 1 to 12 and 1 to 100.

Here are examples of a piece of a vignette scanned by the same device as above.

0 compression (100% quality)	20 compression (80% quality)
40 compression (60% quality)	60 compression (40% quality)
80 compression (20% quality)	100 compression (0% quality)

Let's keep perspective.

Even at this scale, the example with 80 compression is still perfectly usable for the web. The 100 compression image, however, is completely unusable for this degree of enlargement. But that does not mean high compression is without purpose!

Here are two greatly reduced examples of the source certificate used above. The image on the left was originally saved with no compression and the one on the right was saved with maximum compression. Both were then reduced to a mere 200 pixels wide like used for thumbnail images in thousands of web sites. At 200 pixels wide, the image on the left looks a little better, but hardly enough to matter.

100 compression (0% quality)	0 compression (100% quality)

If you are really interested in making web sites responsive, do not be afraid to use very high JPG compression. High compression makes for small file sizes and super-quick web response. Know your purpose, experiment, and scan for your intended needs.

Back to Scanning home page