CCD cameras have been available to amateur astronomers since the late 80's with the introduction of the SBIG ST-4. CCD is the acronym of "Charge Coupled Device". The way a CCD array transforms an image into a computer file is quite simple in principle. Light falling on a grid of detectors produces a pattern of electric charges, which are measured, converted to numbers, and stored in a computer.

The quantum efficiency of a sensor describes its response to different wavelengths of light. Standard front-illuminated sensors are more sensitive to green, red, and infrared wavelengths (in the 500 to 800 nm range) than they are to blue wavelengths (400 - 500 nm). Back-illuminated CCDs have exceptional quantum efficiency compared to front-illuminated CCDs.

Some sensors offer an optional anti-blooming gate designed to bleed off overflow from a saturated pixel. Without this feature, a bright star which has saturated the pixels will cause a vertical streak.

This Anti-blooming gates built into the CCD occupy about 30% of the pixel area. The result is a 70% fill factor and reduced sensitivity and well depth. The reduced sensitivity means that you have to expose almost twice as long to get the same signal level as a CCD without the anti-blooming feature. Also, the area of the CCD occupied by the anti-blooming gate leaves a significant gap between pixels, reducing the effective resolution of the sensor.

Most CCDs have the ability to clock multiple pixel charges in both the horizontal and vertical direction into a single larger charge or "super pixel." This super pixel represents the area of all the individual pixels contributing to the charge. This is referred to as binning. Binning of 1x1 means that the individual pixel is used as is. A binning of 2x2 means that an area of 4 adjacent pixels have been combined into one larger pixel, and so on. In the latter exemple the sensitivity to light has been increased by 4 times (the four pixel contributions), but the resolution of the image has been cut in half.

Digitization, also referred to as analog to digital conversion, is the process by which charge from the CCD is translated into a binary form used by the computer. The term binary refers to the base 2 number system used. A 12 bit camera system will output 2 raised to the 12th power or 4096 levels. A 16 bit camera will output 2 raised to the 16th power or 65536 levels.

Long exposure deep sky images obtained with a CCD are always distorted by three types of defects:

During the exposure, a thermal interference signal builds up in each pixel while collecting the signal from the observed object. The CCD needs to be cooled to a cryogenic temperature (100K, or -173°C) to reduce this signal to a negligible level. Unfortunately, most CCD cameras suffer from the effects of this signal when they are only slightly cooled (temperature over -50°C). The principal consequence of the dark current (whose name comes from the fact that is it present even when the detector is in total darkness) is noise on the image. This noise has two origins:

• reading noise linked to instantaneous fluctuations when the information from each pixel of the CCD is read. The only simple way to reduce this noise is to acquire several images of the same object and then average them;

• spatial noise due to the fact that each pixel in the image reacts differently to the dark current, which gives a grainy aspect to the raw image. This difference in sensitivity to the dark current is strongly correlated from one image to another, so it is possible to produce a reference map of the interference signal to correct the images. The dark current map is obtained by the accumulation of many (typically 7 to 10) long exposure images taken in complete darkness. The CCD should be cooled as usual to reduce the reading noise.

In addition to the useful signal, there is a DC offset that has a value that is independent of the integration time and somewhat independent of the CCD temperature. This offset is induced by the characteristics of the CCD output amplifier and of the electronics of the video signal processor. This signal can easily be eliminated by subtracting an offset map from the images to be processed. This map is obtained by averaging several images acquired with a minimum integration time in total darkness.

The CCD pixels do not all have the same sensitivity to light. As a consequence, if the detector is illuminated with a uniform source, the resulting image will not necessarily be uniform. Just as for the dark current, a spatial noise that degrades the detectability and the photometric quality will be recorded. In addition, the telescope focus is never perfectly uniform because of optical vignetting or the presence of dust in the path of the light beam. These effects are like a local fluctuation in the gain of the CCD. This gain can be corrected for by dividing the images to be treated by the image of a supposedly uniform field. This image, called the flat-field image, is obtained at dusk, when the sky is still bright enough to get a relatively strong signal (between 1000 and 2000 levels) with a short integration time (between 1 and 30 second exposure). If it is too dark, the exposure time will need to be too long, and there will be unwanted stars on the image.

The most basic steps in pre-processing a raw CCD image involve subtracting offset or bias and dark frames and dividing by a flat field frame. The result is a pre-processed or calibrated image. Individual dark and flat fielded frames or calibrated images can be added or averaged for better results.

CCD revolution

Messier 51
Stephan's Quintet
Messier 57

Image restoration