Hello,
I am trying to apply the autocorrelation technique explained in this
article (specifically pages 3 and 4) to compute the impulse response of
interpixel capacitance for a CMOS imager.
Click link to view article:
http://dl.dropbox.com/u/50978256/QE_overestimation_from_IPC.pdfI want to apply this technique to a set of 32-bit noise images that were
created by computing the difference image D[i,j] of two identcal flat-field
frames A1[i,j] and A2[i,j]. In order to compute this impulse response I
need to first determine the autocorrelation function of the difference
images RsubD[x,y] by taking the inverse fourier transform of their spectral
power densitities SsubD(E,n) (page 3). Then the autocorrelation functions
are applied to equation 27 (page 4) to compute the interpixel capacitance
impulse response h[x,y]. I attempted to follow this technique by
correlating a difference image with itself using Process>FFT>FD
Math>Correlate (un-checked the box to perform inverse FFT). I then took
the square root of the FFT produced by the correlation function and then
did an inverse FFT. I am not sure if my method is correct and could really
use some insight on how to properly apply this technique in ImageJ. Here
is exactly what I am unclear about:
1. How to apply ImageJ to compute the autocorrelation function RsubD[x,y]
of my difference images.
2. How to apply ImageJ to compute the 2sigma subN squared term in Equation
27.
Any help would be much appreciated.
Thanks,
-Aaron.
Locked
