Increasing bit depth can allow a scanner to capture color at more than eight bits per channel and transform it to eight bits for output to a file or image application program. Together with good dynamic range, this increased bit depth processing also can improve ability to obtain detail from the darker areas of an image.
Bit depth determines the precision of a scanner's tonal output.
A 24-bit scanner has eight bits per channel (red, green and blue)
and, usually, eight bits for its gray scale channel. With eight bits
per channel, the CCD analog voltage signal from
of an image pixel can be converted to only one of 256 values,
ranging from 0 for the darkest to 255 for the lightest, as
illustrated by upper line of input values in the graph on the left.
A 30-bit scanner can create 10-bit data values per channel. This means that CCD analog input values for each pixel in a channel can be converted to a finer granularity of values: one of 1024 (from 0 darkest to 1023 lightest), illustrated by lower line of input values in the graph on the left. This provides four times as many digital input values from the CCDs values as are available with a 24-bit scanner. A 36-bit scanner would offer even more precision with a range of 4096 input values per channel.
Gray scale image is provided in a single channel that usually has the same input value range as each of the color channels, which is why scanner specs will say 24-bit color 8-bit gray, 30-bit color 10-bit gray, or 36-bit color 12-bit gray. The concepts described here apply individually to each of the three channels for color scans, to their composite representation in a RGB "master channel" and to a single gray scale channel.
Even with increased input bit depth, most scanners will provide output of only 24-bits - eight bits per channel - to a file or image editing software. So how are a greater than eight scanned bits per channel useful when scanner will deliver only eight? A scan bit depth greater than that provided in output can help to obtain better detail in the shadow areas when some tonal shifts are made after the analog to digital conversion. Such tonal shifts are made to compensate for visual nonlinearity and are typically introduced via gamma correction, adjustment of the scan histogram midpoint, and from some other adjustments in the scanner driver software.
A scanner's map of digital input data from its CCDs to digital output data sent to an application or file is usually not a straight line function because of these adjustments, which most commonly exert the transform effect of a power function curve that rises most rapidly in the range of lower input values. For example, the input-to-output map shown above for an input range of 0-255 (eight-bit channel) will map input 63 to output 127. This means that there is a smaller range of input values to spread among a larger number of output values in this dark-area range. For a 10-bit channel the range is 0-1023 and the input value that maps to the 127 'halfway point' in output values is now 255 in the 0-to-1023 input range. It was 63 in the 0-to-255 input range. Thus there can be finer grained input data from the darker region of the image - more input values than output values, which may give the scanner's output mapping function greater detail to work with in for low output values within its eight-bit output range. (Some references say these are "better bits!")
We say here that finer grained input values may yield better detail in the darker areas. This is where dynamic range becomes important. Dynamic range defines how dark the darkest part of a image can be and how light the brightest can be while still showing detail in both. The dynamic range of the source material must actually include the desired detail. A photo print, for instance, can have a maximum contrast range of about 100 to 1; a halftone printed page usually has even less. For some source material there may not be any 'dark vs dark' differentiation within the darkest areas.
The other important dynamic range value is that for the scanner. To be useful, the dark image area pixel data delivered by the additional input bits must represent real data from image, not low level noise from the scanner and its electronics. Increased dynamic range means stronger signal and less noise: a better signal to noise ratio - more likely that low input values represent real scanned image data rather than noise. Conversely, lower values of dynamic range mean that more of the lower input data values may represent noise rather than true image content from the darker areas of the scanned image.
Bottom line: Increased image quality from a good 30 (or 36) bit scanner that also has good dynamic range may be obtained from the combination of greater bit depth and higher dynamic range. Greater bit depth will offer more finer grained input values, allowing a better distribution of output values in the darker areas of images corrected by typical digitally-applied output mapping; and higher dynamic range means that more of those input values may represent image data rather than noise.
Some links for additional information and discussion:
Bit depth - how much do you need? - Tony Sleep's Filmscanner pages
Extended-bit scanning, 30, 36 and 42-bit Scanners (UMAX)
Dynamic range, 24 bit or 30 bit color depth in scanners - Wayne Fulton's scan tips
Bruce Fraser: The High-Bit Advantage --
Why bother working with 36- or 48-bit color when you're stuck with 24-bit output?
Gamma: how it affects image quality (includes examples from many scanners)
CGSD - Gamma Correction Home Page
Charles Poynton's Gamma FAQ and Color FAQ
Content: January 98, July 98; Links: April 2001
Bob Shomler