| Re: Scale the psf on images. [message #90013 is a reply to message #90012] |
Wed, 14 January 2015 23:58   |
Mats Löfdahl
Messages: 263
Registered: January 2012
|
Senior Member |
|
|
Den torsdag 15 januari 2015 kl. 08:13:19 UTC+1 skrev anes.tz...@gmail.com:
> Well perhaps i will get the award for the dumbest person in the world for this.
>
> On the following link there is an fft example
> http://www.exelisvis.com/docs/fftreducebackgroundnoise.html
>
> As you said Mats they do something to reduce the noise. The problem is by using this method, again the final image doesnt make sense. There is no structure actually and the pixel values vary from -1e7 to 1e7. This is the code
>
> fits_read, 'Image1.fits',image1,header1
> fits_read, 'Image1.fits.psf.1.fits',image2,header2
> fits_read, 'Image2.fits',image3,header3
> fits_read, 'Image2.fits.psf.1.fits',image4,header4
>
>
> ;Compute the difference kernel
> p1=fft(image2)
> p2=fft(image4)
>
> ; Compute the power spectrum of the transform and apply a log scale.
>
> powerSpectrum = ABS(p1)^2
> scaledPowerSpect = ALOG10(powerSpectrum)
>
> powerSpectrum2 = ABS(p2)^2
> scaledPowerSpect2 = ALOG10(powerSpectrum2)
>
>
> ; Scale the power spectrum to make its maximum value equal to 0.
> scaledPS = scaledPowerSpect - MAX(scaledPowerSpect)
> scaledPS2 = scaledPowerSpect2 - MAX(scaledPowerSpect2)
>
> surface,powerSpectrum2
>
> mask = REAL_PART(scaledPowerSpect) GT -5
> psf1_trans = p1*mask
> mask2 = REAL_PART(scaledPowerSpect2) GT -5
> psf2_trans = p2*mask2
> kernel = REAL_PART(FFT(psf1_trans/psf2_trans, /INVERSE, /CENTER))
>
>
> image3_prime = convol(image3, kernel)
> diff = image1 - image3_prime
>
>
> WRITEFITS,'im_conv.fits',diff,header3
> end
You are aware that you are trying to construct noise filters based on the power spectra of the PSFs?
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
