Figure 7.2 Reducing quantisation and digitisation artefacts by multi-frame averaging – all pictures are a
zoomed in portion of a full image frame from a dark field image acquisition experiment with the AF51
camera, all taken from the same area and contrast is stretched for the purpose of this illustration. From
left
to
right
: Single frame MJPEG stream saved as a JPEG image (both compression blockiness and
quantisation artefacts are prominent), 64-frame average MJPEG stream saved as a JPEG image
(compression artefact blocks are reduced in size but the range of visible grey levels is still very limited), 64-
frame average from the uncompressed YUYV image stream saved as a BMP image (compression
blockiness is gone – we can resolve single pixels – but the limitation on the number of grey scales, the
quantisation artefact, remains because this image was saved in an 8 bit format, 64-frame average from the
uncompressed YUYV image stream saved as raw doubles showing a much smoother gradation of grey
levels at single pixel resolution – the compression artefacts have been eliminated and the quantisation
artefacts have been greatly reduced (the amount of reduction of quantisation artefact, just like the amount of
reduction in random noise, will increase as we increase the number of frames we average).
Dark current signal, bias and amplifier patterned noise
The presence of a fixed pattern of pixel readout information in total darkness is common to
CCD and CMOS cameras and seldom are pixel values ever 0 even in complete darkness.
For this reason, for quantitative imaging, it is recommended to subtract a master dark
frame from all captured image frames prior to any other processing. The PARD Capture
software allows you to do this this for each image frame as it is captured thereby making
dark frame correction very easy and convenient.
Because on chip amplifier characteristics will change with overall gain and because dark
current characteristics will change with temperature of the chip and exposure time, a
master dark field image (‘master dark’) should be acquired using the same constant
exposure settings as will be used when taking the actual pictures you are going to correct
with that master dark.
As with flat field correction, multi-frame averaging and an uncompressed image stream
should be used when making a master dark and you should save the result in raw doubles
format (or FITS format if you want to use it in some other software). The procedure is
essentially to cap off the lens of the camera to avoid any light getting in and then collect
the multi-frame average image at a particular constant exposure setting. Dark field
correction comes before all other processes so no flat field correction or masking is done
OptArc AF51 Camera Page 90 of 99 User Guide v1.02