At the moment i am working on fixed pattern noise. I need to calibrate
a camera which has an output for the pixel y = (a1)(k1)X + b1 + b2. I
can have a measure value for k1 and b2 with an error factor in
it(assume 5%). I need to produce an estimation algorithm to estimate
gain a1 which can have values from (0.9 - 1.1). b1 need to be
estimated as well. This estimation has to be done for each pixel
separately since each pixel has a different value for gains (a1, k1)
and offsets (b1, b2). Each column of the image array has the same k1
and b2. I would like to produce an algorithm which will start with a
look up table base on off board calibration and since we can measure
k1 and b1 we can give a value for a1 and b1. On the other hand this
algorithm will save its results (the values of k1,  b2 and the
corresponding a1, b1) on a table. The truth is that a1 and k1 are not
correlated but at the time that you switch on the camera the main
affect will on the gains and offsets will be the ambient temperature.
Afterwards the illumination on each pixel will affect these parameters
mostly a1 and b1. I am looking for a way to implement such a thing.