
Thanks for the reply. Yes, I see now that you gave the details earlier in this thread, I was mislead by the "correlation" label. This is not just the correlation of one signal by the negative of ...

Well, based on theory you can easily compare pixel integration with a linear ideal low pass filter. For a given integer decimation ratio they result in the same total noise: but the ideal lowpass ...

I think the fill factor must increase then.

I have seen this marketing talk before, I remember it being from Canon that time. What Canon meant was that the active capture area of a pixel (including the effective fillfactor) was unchanged ...

Yes, zero response occurs when cycles/mm = pixels/mm. That is when there is one full cycle per pixel. Nyquist is two pixels per cycle or 0.5 cycles per pixel.

This is an odd feature. Correlating two signals (one inverted for the difference correlation) does not introduce new frequencies: correlation in the space domain corresponds to multiplication of ...

No, 200% fill factor (or in this case overlapped uniform blur) would have a null at Nyquist, 100% has a null at fsamp. At Nyquist the response is 0.6366.

I was thinking that only the sensors in the middle of the strip would be exactly aligned when in focus, and the ones to either side would be progressively offset in and out, but it looks like you ...

Shouldn't the stagger be mirrored for the left and right sensors so that when the image lines align with the stagger of one sensor they are not aligned with the other sensor?

Correction: the raw shot noise is independent pixel to pixel, so its spectrum *is* white. The expected standard deviation of the shot noise varies pixel to pixel, so if we were to take the Fourier ...

Correction: first convolving with the rectangular pixel integration yields an amplitude of sinc(1)= 0. If the pixel active region was 1/2 the pixel spacing then the value would be sinc(0.5).

No, you didn't do that correctly. Take the 1D case. After sampling we have a train of impulses which represents the sampled signal, then convolving this signal with a rectangular aperture ...

I think that it is really better to consider downsampling as two separate stages even though these are usually combined for computational efficiency (since it isn't efficient to calculate results ...

If you just model sampling as a train of impulses, then you would see replications of the base spectra; this is what interpreting the discrete Fourier transform as the Fourier series coefficients ...

If shooting motion for stills, then of course you would choose your shutter speed to optimize the still images. With the red cameras you can set an HDRX mode so it records a short exposure and...

If shooting motion for stills, then of course you would choose your shutter speed to optimize the still images instead of just using half the frame interval like video usually uses. With the red...

See the second part of my PSF post above to see how to move from the frequency domain back to the space domain. The trick of moving the FFT phase reference using a linear spacial phase shift is ...

The convolution of the diffraction PSF of a square aperture and the pixel integration PSF in the spatial domain can be plotted by simply convolving the two PSF: Nup= 32; Nsamples= Nup*256; x= ...

In the frequency domain it is a convolution with a sequence of impulses: the sampling step looks like it folds the input spectrum so that each multiple of fs lands at zero. See the Matlab fAlias= ...

You cannot interchange the sampling with the integration because the sampling is not a frequency preserving linear filter of the input signals: frequencies above the Nyquist rate get aliased to ...
Activity older than 12 months is not displayed.

DSPographer has not added any gear yet.
Total messages 
2293 
Threads started 
24 
Last post 
1 day ago 
