2 0 H a M e s i l a S t . , N e s h e r 3 6 8 8 5 2 0 , I s r a e l
P O B 2 5 0 0 4 , H a i f a 3 1 2 5 0 0 1 , I s r a e l
T e l : ( + 9 7 2 ) - 7 2 - 2 7 2 3 5 0 0 F a x : ( + 9 7 2 ) - 7 2 - 2 7 2 3 5 1 1
Page no. 32 | Configuration Interface
7.7.2
Defect Pixel Correction
The defected pixel correction will correct up to 32 pixels in the sensor and up to 2 adjacent pixels in a row. The
pixel correction coordinates represent pixels of sensor’s visible ROI, therefore identifying the correct X and Y
coordinate should be done using default, full resolution image.
The algorithm will correct the defect pixel based on the value of existing adjacent pixels. The correction for Mono
and Color sensor is slightly different and described as follows:
Mono pixel correction:
The defect pixel
𝑃(𝑥, 𝑦)
value will be the average value of 2 pixels adjacent to pixel
𝑃(𝑥, 𝑦)
from both sides in
the same row.
Figure 29 – Defect pixel correction position for Mono image
Color (Bayer) pixel correction:
The defect pixel
𝑃(𝑥, 𝑦)
value will be the average value of two pixels
from both sides of
pixel
𝑃(𝑥, 𝑦)
in the
same row,
corresponding to the same Bayer color element.
Figure 30 – Defect pixel correction position for Color image
7.7.3
Field Correction
The Flat-field and Dark-field corrections are used to improve the quality of the image by removing the artefacts
that are caused by fixed pattern noise and variations in the pixel-to-pixel sensitivity of the detector.
To make Dark/Flat field (DSNU / PRNU) correction, two pictures should be taken. The Dark/Flat field correction
algorithm operates on rows/columns of the sensor and therefore is unable to correct pixel wise distortions like
lens shading.
For DSNU a reference image has to be recorded in dark, with lens closed (offset should be boosted) or fully
removed from the camera and covered with a solid cap. To calibrate for PRNU a reference image has to be
recorded with a uniform illumination of around 40%. These two separate steps in the flat field correction are
therefore referred to as dark field calibration and bright field calibration, respectively.
The operator is per pixel and defined according to following formula:
𝑃(𝑥, 𝑦) = 𝐺𝑎𝑖𝑛(𝑥)𝐺𝑎𝑖𝑛(𝑦)[𝑃(𝑥, 𝑦) − 𝑃
𝑑𝑎𝑟𝑘
(𝑥) − 𝑃
𝑑𝑎𝑟𝑘
(𝑦)]
