Page 292 - Fundamentals of Light Microscopy and Electronic Imaging

IMAGING PERFORMANCE OF A CCD DETECTOR 275

2
capacity depends on pixel size (holding capacity is 1000 e / m ), CCD imagers with
large pixels usually have a higher dynamic range. For example, a camera with pixels 27

m on a side, a 580,000 e full well capacity, and 9 e read noise, has a DR 65,000.
The read noise of a camera is important in determining the dynamic range as well. Due

to continual improvements in reducing readout noise (down to 5 e and lower for cam-
eras used in biological laboratories), even interline cameras with small 4.5 m pixels
can have a DR 4000.
To realize the potential dynamic range of a CCD camera, the camera must be
equipped with an appropriate digitizer, also called an ADC or digital processor. Digitiz-
ers are described in terms of their bit depth. Since a computer bit can only be in one of
two possible states (0 or 1), the bit depth is described as 2 number of steps. Therefore,
X
8, 10, 12, and 14 bit processors can encode 2 . . . etc. steps, or a maximum of 256,
8
1024, 4096, or 65,536 gray levels. A manufacturer would typically fit a camera with a
DR 4000 with a 12 bit processor to match the DR of the imaging system.
On some of the latest interline CCD cameras, the calculated dynamic range is about
2000, so the use of a 12 bit ADC seems to be more than is required. However, some of
the new designs make full use of the 12 bit digitization by including a ⁄ gain setting.
1
2
This option is included because CCD chips are designed such that the pixels of the serial
register hold twice as many photoelectrons as pixels in the parallel register. Thus, when
a camera is operated with 2 2 binning, a common mode of operation for routine
microscopy, 12-bit imaging at high quality can be obtained.
It is now quite common to see cameras fitted with processors having a much higher
digitizing capacity than the inherent DR of the camera. If the read noise is designated as
1 electronic gain (the conventional assignment for gain), there are potentially a large
number of unused intensity levels when the bit depth of the processor exceeds the
dynamic range of the camera. To fill these extra levels, manufacturers apply a (hidden)
2–4 gain to boost the signal to utilize the gray levels of the processor, but as we have
seen, this operation increases the noise in the image. Therefore, by noting the full well
capacity of the pixel and read noise of the camera, the dynamic range of different cam-
eras can always be calculated and compared.
In comparison to high-bit CCD cameras, display devices such as computer moni-
tors and dye-sublimation printers use only 8 bit processing (256 gray levels). In fact, the
dynamic range of communications media such as TV/video, pictures in newsprint, and
even of the eye is a meager 5–7 bits, depending on the image and illumination condi-
tions. If the visual requirements for printers and monitors are so low, why are high-bit
imaging systems necessary or even desirable? High-bit depth (DR) is required:

• For purposes of accurate quantitation of light intensities, for example, for examin-
ing ratiometric or kinetic imaging data; the larger the number of gray levels, the
more accurately an intensity can be described.
• To perform multiple image processing operations without degrading the image;
because of mathematical round-off errors during processing, images with more
gray levels can tolerate a greater level of mathematical manipulation.
• For selecting and working with a portion of an image for display, where the region
of interest covers only a narrow portion of the dynamic range of the full image. For
an ROI including just 2% of the full dynamic range, this would be 82 levels for a 12
bit system, but only 5 levels for an 8 bit system. If the 5 gray levels were now
stretched to fill the 256 levels for an 8 bit monitor or print, the image would look
pixelated, with noticeable steps or contour lines in it.