[Elphel-support] linescan & artefacts

[Michael Aschauer]

Andrey,

I read up a bit and did few experiments and it looks like it should be 
solvable with a flat field correction (image diveded through a 
normalized grey card measurement picture) combined with darkframe and 
bias frame substraction...I do not have a useful workflow for that yet...

However I think a (calibrateable) in-camera/in-FPGA correction 
possibility would not be a bad idea (especially if the code is already 
there)

best,
m,-


>
>     the pattern changes depending on which line-pair is used. but the
>     output is not necessarily any better...
>
> I see. So it looks like pixel fixed-pattern noise. There are two ways
> how to fight it.
>
>   One is to resurrect another part of unused FPGA code - the one that
> can subtract/multiply the fixed values from the image pixels (that will
> also need some calibration).
>   Second - with post-processing. In that case you should rather use jp46
> format where pixels are not interpolated. First step - eliminate
> additive parameter - you can get it if you average multiple line pairs
> of the grey card or (even better) - disable autoexposure and white
> balance (if you use them) and close the lens. For such measurements you
> need to un-apply gamma (you can do it in ImageJ plugin) so the
> background subtraction will take place with linear pixel values (setting
> gamma=1 in the camera would kill the dynamic range)
>
> Andrey