5 –Camera Controller
A6700sc/A6750sc User’s Manual
30
value directly relates to the gain coefficient in the NUC table, the A6700sc can scan the NUC table
gain coefficients and use them to determine if a pixel’s responsivity exceeds the limits as set by the
user.
The second method of determining bad pixels is to search for twinklers. Twinklers are pixels that
have responsivity values within normal tolerances, but still exhibit large swings for small input
changes. These pixels are on the “verge” of being bad and often appear to be noisy. To find these
types of pixels the camera collects N number of frames and records the maximum and minimum
values across that sample set for each pixel. If the delta between max and min exceeds the
Twinkler
max pixel value delta
then the pixel is determined to be bad.
Since the responsivity test requires a gain coefficient, it is useless on NUC tables determine by the
One-Point Correction because those tables have a value of one (“1”) as the gain coefficients. The
Twinkler test can be done on either correction process.
The A6700sc uses two algorithms for bad pixel replacement: 2-point Gradient, and Nearest Neighbor.
The 2-point gradient algorithm is the default bad pixel correction method. With this algorithm, the two
pairs of pixels above and below and to the left and right of the bad pixel are evaluated. The algorithm
compares the differences between the pixels and chooses the pair with smallest gradient (difference).
It then averages the two adjacent pixels and uses that value for the replacement value. This
algorithm is better at handling bad pixels near a high contrast edge and is the default method. If the
algorithm encounters a situation is cannot solve (for example, an edge or corner) it will fall back on the
nearest neighbor algorithm.
Nearest neighbor uses a simple replacement using an adjacent pixel. The adjacent pixel is picked
using the pattern depicted below. When a bad pixel is near an edge, those search positions are
skipped.
