CCDBand-Aid repairs vertical bars in Kodak KAI-11000M images

[georg.viehoever]

...
I think that i'm going to write a solution pretty close to the canon banding script. So, if you have any suggestions on that field, I'm all ears.

What I always wanted to try is to enable CanonBanding reduction on RAWs, running it seperately on each of the 4 subpixel sets (R+G1+G2+B). An ability to switch the direction from horizontal to vertical would be nice. When doing a module, a proper realtime preview would be on my wish list.

I am not sure how much CanonBanding reduction will help with the problem observed by Bob. It is very subtle, and much less prominent than the banding you see in DSLR images.

Georg

[Carlos Milovic]

Hi Georg

What I was considering was this: First, calculate the median of each raw or column, and make a vector of these values. Then, take the fourier transform of that vector, and supress high frequencies (I may use the code of the optimal notch filter I'm thinking of). After the inverse FFT, calculate the differences to the old vector, and use that as an additive value for the complete image.
This sounds fast to implement, but I doubt that a RTP would be accurate...

I hope that this procedure would be more robust to artifacts than a 2D FFT filtering process.

BTW, I may also try to use a "training" image, to calculate the correction, and then apply the difference vector to the target image, so it may yield better results in Bob's problem.

[georg.viehoever]

Carlos,
Bob also said that the pattern had a square waveform. I am not so sure that FFT techniques would work well for these. But by all means, give it a try!
Georg