66
•
Camera Operation
Flat Field Correction and Defective Pixel Detection Overview
The Flat Field correction function consists of using two coefficients per pixel which correct
the gain and offset of the corresponding pixel. These corrections compensate for the Photo-
response Non-uniformity (PRNU) and Fixed Pattern noise (FPN) attributes unique to each
camera sensor. In addition, the camera supports replacement of defective pixels (hot, dead,
blinking) with a value based on neighborhood pixels.
Correction Function Block Diagram
The following simplified block diagram shows the processing chain that is applied to the
image data (the flat field and defective pixel blocks are highlighted). Note that each
processing block can be activated and deactivated independently. For example, the FPN and
PRNU coefficients can be applied independently or together using the
flatfieldCorrectionMode.
Figure 21 Flat field and defective pixel processing
Flat Field Correction Algorithm Description
Flat Field Correction Algorithm–Method1 (feature:
flatfieldCorrectionAlgorithm
) applies the
following FFC formula for correcting pixel values:
newPixelValue
x,y
= (sensorPixelValue
x,y
– FFCOffset
x,y
) * FFCGain
x,y
where:
•
x
&
y
are the Flat Field Correction Pixel coordinates.
•
See the flatfieldCorrectionPixelXCoordinate and flatfieldCorrectionPixelYCoordinate
features.
•
newPixelValue
is the pixel value after Flat Field Correction is applied.
•
sensorPixelValue
is the pixel value before Flat Field correction is applied.
•
FFCOffset
is the offset coefficient value to subtract from the sensorPixelValue.
•
FFCGain
is the gain coefficient value that is multiplied with the sensorPixelValue.
The implementation of this formula requires that both the FPN and PRNU coefficient are
stored in 16 bits. For the Falcon2 we reserve 7 bits for the FFCOffset (FPN) coefficient and 9
bits for the FFCGain (PRNU) coefficient. The FFCGain can be calculated as follows:
FFCGain
x,y
= (FFCGainRaw
x,y
/ GainDivisor) + 1.0
where:
•
x
&
y
are the Flat Field Correction Pixel coordinates.
•
FFCGain
is the floating point multiplier of the sensorPixelValue.
•
FFCGainRaw
is the stored 9 bit value representing the FFC gain value.
