The first source of this additional signal is known as dark current and is a result of signal accumulation on the detector from thermal conditions. The dark current of the MSI was analyzed by acquiring many images during piecepart and full instrument testing without any illumination. Any observed signal in these images above the fixed floor offset of the instrument is attributable to the dark current. Since this signal is typically not interesting when doing photometric analysis of the images, the dark current must be subtracted prior to performing the other calibration steps.
Even/Odd Pixel Variation
There is another problem inherent in all of the MSI images that
should also logically be removed at the same time as the dark current.
This problem is a variation of DN values between all of the even
pixels and all of the odd pixels on the order of 5 DN. Figure 1 shows a heavily magnified and contrast
enhanced portion of an MSI image that clearly shows the even/odd
variation.
Figure 1
Because this variation is independent of signal, its effect is
included in all of the dark current images. Since the proper handling
of this variation is to subtract it from the original images, then
subtracting the raw dark current images from signal images will also
remove the even/odd variation in a single step.
Correction Algorithm
Since an exact numerical formula for dark current and even/odd
variation will never be known, the approach that was selected uses
existing dark current images taken at various exposure times and
temperatures to generate representative images of the dark current
that are subtracted from the signal image.
Mark Robinson has devised this plan based on linear behavior of the dark current as a function of both exposure and temperature. His IDL script dc_interp.pro takes a list of dark current images acquired at the same temperature as the original image and computes an average dark current image for the exposure time of the original image as a linearly interpolation of the input dark images.
Because the current number of dark current images that are available for each temperature and exposure condition are somewhat limited, a single average value is computed for all of the odd pixels on a line, and a single average value is computed for all of the even pixels on a line. These two numbers are subtracted from the even and odd pixels respectively of the original image. This process is repeated for each line of the image.
If a large sampling of dark current images are acquired during cruise at a variety of temperature and exposure conditions, a set of pixel by pixel average images could be computed for performing the dark current and even/odd removal.