Scanning your photos - tutorial

What does a scanner do?

A scanner breaks the original down into tiny dots and measures their brightness. However, even color scanners are, paradoxically, color-blind! In order to still be able to generate colorful scans, scanners use a trick: by inserting color filters in the primary colors red, green and blue, the respective color components are measured separately. If you superimpose these three color separations, which the scanning software naturally does automatically, you get a color image whose elements are defined by mixtures of these three primary colors. However, each one of these individual dots has a homogenous color, meaning that the color doesn't vary within one dot. These virtual "mosaic tiles" (you could define them as the atoms of the digital image) are referred to as dots or pixels (contraction of picture elements). The exciting thing about digital imaging in general is the fact that you no longer deal with the image as whole, but rather with its individual components. The software enables the image editor to work with each picture element individually, like changing its color, brightness or position, or copying its characteristics to adjacent picture elements.

Resolution

Not all dots are created equal

Sample depth and color resolution

Five hard facts about scanning

Resolution

The term resolution is closely related to the subject of breaking down a paper print or slide into pixels. Resolution is the magic word these days and crops up in every kind of communication on the subject of digital images, scanners or printers.

dpi and ppi

dpi stands for "dots per inch". It actually refers to the dots that a printer can print on paper. In contrast, ppi is the abbreviation for "pixel (or points) per inch". The following paragraphs will clearly explain why printer dots are not the same as image pixels. Although ppi is actually the more correct term for the resolution of a scanner - a scanner does, after all, divide an image into pixels and not printer dots - the abbreviation dpi has also become an established term to refer to resolution. For example, you will come across dpi in scanning software everywhere where you have to set the scanning resolution, but also in advertising, the trade press and the various image editing programs. In order to avoid confusion, this chapter also uses the term dpi.

For example, if a scanner has a resolution of 600 dpi, this means nothing more than that its scan array can divide one inch of the original into 600 elements. To do so, the scanner has a corresponding number of light-sensitive receptors, each of which measures the brightness of a tiny section of the original. Consequently, a 600 dpi scanner is capable of dividing 2.54 centimeters of the original (1 inch = 2.54 centimeters) into a maximum of 600 pixels.

It's the maximum because you naturally don't always have to scan at "full power". Therefore, the scan resolution can be reduced as needed using the software, so that one inch is only divided into 300, 200 or 72 dpi, for example.

Optical and interpolated resolution
A distinction must be made between two terms when referring to the resolution of a scanner. The optical (or physical) resolution refers to the actual number of photocells per unit length. In contrast, the interpolated (or calculated, or maximum) resolution is the result of a kind of software trick. Based on the optical resolution, the scanning software generates additional pixels, calculated by finding the mean, and inserts them in between the pixels actually scanned. Used in moderation, you can thus increase the amount of data. However, it must be kept in mind that interpolation does not lead to the addition of really new image data. In other words, an optical resolution of 600 or 1200 dpi, for example, generates better results than an interpolated resolution.

The scan resolution thus defines how many pixels one inch of the original is divided into. As will be illustrated later on, this has an effect on the file size and also on the dimensions of the scan.

Not all dots are created equal

The resolution of a scanner, but also that of a printer, is often given in dpi, and this is where things start to get a little complicated. The main culprit behind the confusion on the subject of resolution is the fact that the term resolution and the abbreviation dpi mean something completely different for a scanner (and therefore the correct unit should be ppi instead of dpi) than they do for a printer, and that not all dots are created equal!

The reason is the different way they handle color. A scanner with 600 dpi generates 600 pixels per inch, each of which can have one of 16.8 million colors.

A printer (ink-jet, dot-matrix, laser) with a resolution of 600 dpi prints a maximum of 600 dots (individual drops of ink or toner dots) on one inch of paper, each of which can be either:

One of two colors (black-and-white printer, namely black or the color of the paper),
One of four colors (printers with three inks),
One of five colors (printers with four inks),
but never one of 16.8 million colors.

This already clearly illustrates that a scanner "dot" (which, as described above, should correctly be referred to as a pixel) in no way corresponds to a printer "dot". In order to avoid insulting our eyes with poster-like two-tone photos, the printer must therefore simulate each individual pixel of the scanned photo (i.e. scanner dpi) by juxtaposing and overlapping numerous printer dots (i.e. printer dpi) in the primary colors. Due to the fact that the printer "consumes" numerous printer dots in order to render a single pixel, its effective print resolution is substantially lower than the specified dpi value.

It is well known, however, that there are also printers which produce better photo quality with just 300 dpi than high-resolution ink-jet printers with 720 or even 1440 dpi. The reason is simple: thermosublimation printers are capable of printing one of 16.8 million colors for each individual dot of the digital image, which is why they are also called "continuous tone" printers. Thus, a pixel really does correspond to a printer dot in this case.

The following must be kept in mind when scanning:

If the image is to be output on an ink-jet or laser printer, it doesn't help much to set the scan or image resolution to the value of the printer resolution. It you really wanted to do so, even a print size of 9 x 13 centimeters on a 1440 dpi printer would require a data volume of 107 megabytes! That would be going a little bit too far because, as described, the printer is incapable of producing 1440 pixels per inch. The effective printer resolution varies from model to model and can only be determined more precisely by experiment. However, you'll usually be on the safe side with a scan or image resolution of 300 dpi. Less could impair the print quality and more would not improve it.

In contrast, things are easier if the image is to be sent to a continuous tone printer (e.g. thermosublimation or donor printer) or produced as a real photo by a digital-photochemical process: set the image resolution to the same value as the printer resolution, which is usually around 200, 300 or 400 dpi, depending on the model.
How much is too much?

It is generally true that the higher the resolution of a scanner, the more details it can read off the original. However, there are natural restrictions in this case, too. For example, it makes little sense to scan a tiny original, like a passport photo, at a huge resolution, like 5000 dpi. The amount of information the scanner is theoretically capable of recording just doesn't exist on the small original. Therefore, scanners with a resolution of 600 or 1200 dpi are almost always adequate for everyday use.

The color of each dot in the digital image can be defined by indicating its red, green and blue value. One byte of memory is available for each primary color. In turn, one byte corresponds to eight bits and a bit can be equal to either 0 or 1. This storage capacity enables the definition of exactly 256 different shades: from 00000000 to 10000000, 11000000, 10100000, etc. up to 11111111. In other words, one byte or eight bits per primary color, can describe 256 shades of that color. And because a pixel consists of three primary color values, as explained, this description language results in:

256 x 256 x 256 = 16,777,216 or about 16.8 million possible colors.

This gamut of colors is logically referred to as "true color". The corresponding devices, such as scanners, digital cameras or graphic cards, have a sample depth of 3 x 8 bits = 24 bits.

If all the values for red, green and blue are set to zero, they result in black, while 255, 255, 255 produces white. If the three color proportions are equal (such as 150, 150, 150), they define a neutral gray.

Those of you who've held out up to this point now have an easy opportunity to discover another mystery behind the digital image: the relationship between the image dimensions, resolution and file size. The best way to illustrate this relationship is by an example: Scanning a 10 x 15 cm photo at a resolution of 300 dpi results in (reminder: 1 inch = 2.54 centimeters):

(15 centimeters/2.54 cm) x 300 = 1771 pixels horizontal
(10 centimeters/2.54 cm) x 300 = 1181 pixels vertical

Thus, the entire scan consists of: 1771 x 1181 = 2.1 million pixels.

Due to the fact that, as described above, each pixel requires 3 bytes of memory for the color data, the file consists of roughly 6.3 million bytes. If you divide this value by 1024, you get the size in kilobytes (6130) and if you divide this by 1024 again, you get the customary file size in megabytes (6).

Sample depth-crazy?
Now, if the monitor, the image editing software and the photo-quality printer can work with a palette of 24 bits or 16.8 million colors at best, why do high-quality scanners boast about sample depths of 30 or even 36 bits? The reason is simple: the greater the sample depth, the better the scanner can distinguish between fine graduations in the shadows and highlights. It has one billion shades at its disposal at 30 bits, and even 68 billion at 36 bits. Although this richness of color is ultimately reduced back down to the common 24 bits for further processing, the critical areas are still rendered with much greater differentiation due to the more detailed input.

Dynamic range
The value of the dynamic or tonal range also says something about the quality of high-end professional scanners, in particular. On film material, the tonal values are usually measured by the optical density. They theoretically range from 0D (clear, transparent film) to 4D (black, fully exposed film), but are usually around 0.2 to 3.8 in practice. When it comes to scanners, the dynamic range, which is calculated using a mathematical formula, defines the capability to distinguish between a specific range of the entire spectrum from 0 to 4. It can generally be said that the greater the dynamic range, the better. Unfortunately, this usually also means considerably higher prices.

Five hard facts about scanning

You can use a scanner to make a blurry photo sharp? False, because the GIGO rule applies to scanning too, and this rule says: Garbage In, Garbage Out. Although scanning and image editing software offers various sharpening filters, they primarily serve to reduce the slight blur caused by scanning. No creative tricks in the world can turn a really blurry original into a sharply focussed picture. The lesson to be learned is that the optimization functions of a scanner are no substitute for good photography. Even the best scanner can't recreate image details which can no longer be made out in the original, like highly overexposed areas. The higher the resolution the better? False, because there is a right resolution for every application. A low resolution is not suitable for printing a poster, while a high resolution would not only be superfluous for displaying an image on an Internet page, but even problematic. The image would require long transmission times and not fit completely on the screen. You need a very high resolution to scan color photos, while a low resolution is adequate for scanning black-and-white line art? False, on the contrary! You need a high scan resolution in order to reproduce line art, graphics or text without staircasing (jaggies). In contrast, photos usually have varied content without straight lines and edges, so that a low resolution is not so noticeable. If I want to print a picture on my 720 dpi printer in optimum quality, I also have to scan it at 720 dpi? False, because the 720 dpi printer can't really print 720 pixels per inch - see Not all dots are created equal! Scan first and take care of optimization later? False, because optimization steps taken during the scanning process, such as contrast correction, sharpening, descreening or sizing, usually produce better quality results, since they can refer to a greater amount of data (such as the 36-bit raw data) than a standard image editing program.

(*Author: AGFA)