Falcon 4M Camera Manual
03-032-20044-01
DALSA
42
3.7 Flat Field Correction
This camera has the ability to calculate correction coefficients in order to remove non-
uniformity in the image. This video correction operates on a pixel-by-pixel basis and
implements a two point correction for each pixel. This correction can reduce or eliminate
image distortion caused by the following factors:
•
Fixed Pattern Noise (FPN)
•
Photo Response Non Uniformity (PRNU)
•
Lens and light source non-uniformity
Correction is implemented such that for each pixel:
V
output
=[(V
input
- FPN( pixel ) - digital offset) * PRNU(pixel) – Background Subtract] x System Gain
where V
output
=
digital output pixel value
V
input
=
digital input pixel value from the sensor
PRNU( pixel)
=
PRNU correction coefficient for this pixel
FPN( pixel )
=
FPN correction coefficient for this pixel
Background Subtract
=
background subtract value
System Gain
=
digital gain value
The algorithm is performed in two steps. The fixed offset (FPN) is determined first by
performing a calculation without any light. This calibration determines exactly how much
offset to subtract per pixel in order to obtain flat output when the sensor is not exposed.
The white light (PRNU) calibration is performed next to determine the multiplication
factors required to bring each pixel to the required value (target) for flat, white output.
Video output is set slightly above the brightest pixel (depending on offset subtracted).
It is important to do the FPN correction first. Results of the FPN correction are used in the
PRNU procedure. We recommend that you repeat the correction when a temperature
change greater than 10°C occurs (the factory temp is about 37°C,
vt
command). In
snapshot mode 1, FPN coefficients are not particularly sensitive to changes in frame rate
or integration time. In snapshot modes 0 and 2, FPN coefficients will be sensitive to
changes in frame rate.
PRNU correction requires a clean, white reference. The quality of this reference is
important for proper calibration. White paper is often not sufficient because the grain in
the white paper will distort the correction. White plastic or white ceramic will lead to
better balancing.
For best results, ensure that:
1)
60 Hz ambient light flicker is sufficiently low not to affect camera performance and
calibration results.
2)
The average pixel should be at least 25% below the target output. If the target is too
close, then some pixels may not be able to reach full swing (1023 DN) due to
correction applied by the camera.
Note: If your
illumination or white
reference does not
extend the full field of
view of the camera,
the camera will send a
warning.