Ultrasonic sensor jiggling for AA effect

[BJL]

Here is a question for the signal processing experts. One method used to avoid aliasing in CT scanning is to jiggle the sensor slightly, smearing the signal over a few adjacent photo-sites. There seem to be some related patents for photographic applications at HP and Canon.

My naive idea: use the ultra-sonic vibration system of the sensor dust removal system, at lower amplitude: could suitable jiggling of the sensor produce a useful low-pass filtering effect? Especially given Guillermo's observation that photo-sites with high fill factor are already quite good at filtering spatial frequencies that are far enough above Nyquist, so this might only need to have a narrow "absorption band."

The appeal of course is that this could be turned on and off according to the situation: just flip the E/non-E switch.

[BartvanderWolf]

Here is a question for the signal processing experts. One method used to avoid aliasing in CT scanning is to jiggle the sensor slightly, smearing the signal over a few adjacent photo-sites. There seem to be some related patents for photographic applications at HP and Canon.

My naive idea: use the ultra-sonic vibration system of the sensor dust removal system, at lower amplitude: could suitable jiggling of the sensor produce a useful low-pass filtering effect? Especially given Guillermo's observation that photo-sites with high fill factor are already quite good at filtering spatial frequencies that are far enough above Nyquist, so this might only need to have a narrow "absorption band."

The appeal of course is that this could be turned on and off according to the situation: just flip the E/non-E switch.

Hi BJL,

Unfortunately it won't work, because the OLPFs don't shake. Only the IR filter in front of the OLPFs  shakes, probably to reduce mass.

Cheers,
Bart

[BJL]

Hi BJL,

Unfortunately it won't work, because the OLPFs don't shake. Only the IR filter in front of the OLPFs  shakes, probably to reduce mass.

Thanks. Well, so much for naive idea number 1, but it there still hope for the rest of my idea, but instead using a separate, low amplitude, high fequency wobbling of the sensor itself? Little circles of diameter similar to pixel pitch for example? I imagine that the anti-dust shake is far more substantial, given that I can feel it every time I turn my camera on.

[Fine_Art]

You can always Vaseline your lens.

That was the old soft portrait technique.

[BJL]

You can always Vaseline your lens.

Or how about "changing the focusing point or shooting distance"? That is not a joke or nostalgia, it is a current recommendation for handling false color and moiré on the D800E from Nikon itself at http://imaging.nikon.com/lineup/dslr/d800/features01.htm

Thinking about it more, the effect of such jiggling might be no better than increasing the size of the Airy disk by stopping down, though without the reduction in light gathering speed that stopping down would cause.

[madmanchan]

The issue with changing focus slightly is that with most lenses you will show the effects of axial color aberration (greenish or magenta-ish color depending on which direction you defocus).  This is not a problem if you stop down far enough.  Then again, if you stop down far enough, you will get a low-pass effect from diffraction anyways.

[JohnTodd]

Isn't this effectively what happens with the Hassleblad multi-step process, where they shift the sensor less than one pixel-width to increase resolution?

[BJL]

Isn't this effectively what happens with the Hassleblad multi-step process, where they shift the sensor less than one pixel-width to increase resolution?

That reads at each sensor position, with a distinct pause between each readout, and then processes the data to get higher resolution that from a single shot with unmoved sensors. My idea would be akin to then blindly averaging the data from all those positions ... and also doing everything far faster.

[hjulenissen]

Isn't this effectively what happens with the Hassleblad multi-step process, where they shift the sensor less than one pixel-width to increase resolution?

Super resolution is about getting more apparent resolution from multiple, aliased images.

Anti-aliasing is about decreasing the apparent resolution so as to decrease aliasing artifacts.

Hasselblad probably (?) do multiple exposures in their process. The suggestion of BJL probably was to smear a single exposure.

-h

[AJSJones]

Jiggling a filter with no electronic connections sounds simpler than jiggling a sensor with its connections.
Nikon has a patent for an OLPF that can be turned on an off.  It's like the OLPF in the D800E (the one "without" which actually has the components, but with the second one reversing the effect of the first).  In the patent, the second LPF can be in the "on" position  for "normal" AA function (e.g. horiz if the first is vert) or be rotated to a "cancelling" position for the "off" mode.  The rotation is mechanical and adds complexity, however.

[BJL]

Jiggling a filter with no electronic connections sounds simpler than jiggling a sensor with its connections.

True, but sensors are already jiggled (at higher amplitude, lower frequency) by sensor-based image stabilization systems, in Sony and Olympus cameras, for example.

[Guillermo Luijk]

using a separate, low amplitude, high fequency wobbling of the sensor itself?.

Stabilized bodies such as those from Olympus could implement this idea to deliberately un-stabilize the image in order to get an AA effect

[deejjjaaaa]

Stabilized bodies such as those from Olympus could implement this idea to deliberately un-stabilize the image in order to get an AA effect

but it will require redesign of their tech, as currently IS is compensating relatively low freq shake... here you need to do really high freq mechanical movements in 2 directions w/ lower amplitude (sub pixel) while at the same time stabilizing in 2-5 axis' for low freq moves w/ higher amplitude

[marcmccalmont]

The new Nikon D800/D800E could be the same camera with the "vibrating" AA filter turned on or off
Great idea
Marc