Photon Focus MV2-D1280-640, User Manual, Page: 32 / 85

4 Functionality
In robot applications, the stress that is applied to the CameraLink cable is especially high due to
the fast and repeated movements of the robot arm. For such applications, special drag chain
capable cables are available.
4.4 Image Correction
4.4.1 Overview
The MV2-D1280-640 camera possesses image pre-processing features that compensate for
non-uniformities caused by the sensor, the lens or the illumination. This method of improving
the image quality is generally known as ’Fixed Pattern Noise (FPN) Correction’, ’Shading
Correction’ or ’Flat Field Correction’ and consists of a combination of offset correction and gain
correction.
Since the correction is performed in hardware, there is no performance limita-
tion.
The offset correction subtracts a configurable positive or negative value from the live image
and thus reduces the fixed pattern noise of the CMOS sensor. The gain correction can be used
to flatten uneven illumination or to compensate shading effects of a lens. Both offset and gain
correction work on a pixel-per-pixel basis, i.e. every pixel is corrected separately. For the
correction, a black reference and a grey reference image are required. Then, the correction
values are determined automatically in the camera.
Do not set any reference images when gain or LUT is enabled!
Correction values of both reference images can be saved into the internal flash memory, but
this overwrites the factory presets. Then the reference images that were preset in production
cannot be restored anymore.
4.4.2 Offset Correction
The offset correction is based on a black reference image, which is taken at no illumination
(e.g. lens aperture completely closed). The black reference image contains the fixed-pattern
noise of the sensor, which can be subtracted from the live images in order to minimise the
static noise.
Offset correction algorithm
After configuring the camera with a black reference image, the camera is ready to apply the
offset correction:
1. Determine the average value of the black reference image.
2. Subtract the black reference image from the average value.
3. Mark pixels that have a grey level higher than 63 DN (@ 8 bit) as hot pixels that will not be
corrected.
4. Store the result in the camera as the offset correction matrix.
5. During image acquisition, subtract the correction matrix from the acquired image.
30