Patent: Using image sensor shift to reduce the need for an AA filter

[Canon Rumors Guy]

Northlight Images has uncovered an interesting patent that deals with reducing the need for an AA filter.
This patent shows a technique to use the sensor stabilization system to address the problems of sampling and spatial aliasing, something that an AA filter is usually responsible for.
This technology in this patent is for both stills and video, I will note, that understanding this patent isn’t for the faint of heart.
Check out the full patent at the USPTO website.
Interestingly, Canon Rumors reader Nicolas let me know on Twitter that Pentax does something similar. You can check out the demonstration from the Ricoh YouTube channel below.

Continue reading...

 

[dilbert]

Does this mean the camera needs to move the sensor between all 4 of the positions within the time it takes to expose the shot? (A 1/100 shot would mean each of the 4 pixels is exposed as 1/400). Or does it mean that a 1/100 shot now takes 1/25 to complete?

 

[BeenThere]

Would this introduce less blur than AA filter? Obviously more complicated for what benefit?

 

[Sharlin]

Would this introduce less blur than AA filter? Obviously more complicated for what benefit?

You could switch it on and off, for one. Besides, you need the sensor shift tech anyway if you want to do superresolution.

 

[MiJax]

Seems like a lot of effort to simply skin a cat differently.

 

[privatebydesign]

Seems like a lot of effort to simply skin a cat differently.

But that is what companies that sell tech do! Who needs a knife and a stake when you can be sold a multimillion dollar machine and a service contract?

 

[pape2]

Not same what AA filter ,you can turn this off too

 

[notsure]

Pentax have similar tech in their newer cameras, though it only relies on IBIS for this, without cooperating with the lens.

 

[Antono Refa]

Seems like a lot of effort to simply skin a cat differently.

Point is, once sensor shift is possible, you can do different things with it - IBIS, super resolution, AA filter, etc.

If one is shooting star trails, maybe the sensor can be shifted one pixel diagonally between each two shots to cover for dead / hot pixels. If shooting a static scene, it could shoot the scene three times, shifting the sensor to cover each pixel location with the three bands skipping the need for Bayer interpolation.

 

[notsure]

Point is, once sensor shift is possible, you can do different things with it - IBIS, super resolution, AA filter, etc.

If one is shooting star trails, maybe the sensor can be shifted one pixel diagonally between each two shots to cover for dead / hot pixels. If shooting a static scene, it could shoot the scene three times, shifting the sensor to cover each pixel location with the three bands skipping the need for Bayer interpolation.

they actually use the sensor shift as a less-capable astrotracker as well, but their cameras are pretty much made for landscape photography

 

[cgc]

Excerpt from patent's page 9:

"...when calculating the phase difference for focus detection, the influence of its spatial aliasing cannot be ignored, and in some cases, the obtained result of the focus detection may be largely erroneous. Therefore, conventionally, it is common to arrange the optical low-pass filter so that the response in the high frequency region higher than the Nyquist frequency is reduced..."

That is: the aliasing also causes problems to phase autofocus accuracy. So, contrary to some misconceptions, an AA filter is not an obstacle for the AF operation, but a "conventional" requisite to improve its actual performance. A bit of homogeneous blur doesn't mistake the AF system, which is even more fooled by the false detail. AA-free sensors are aliasing machines (nature shots have no less aliasing than its architecture counterparts: simply, our brain accepts as "sharpness" the part of random false detail).

 

[Ozarker]

Seems like a lot of effort to simply skin a cat differently.

Beautiful photos on your Flickr.

 

[Antono Refa]

they actually use the sensor shift as a less-capable astrotracker as well, but their cameras are pretty much made for landscape photography

Who is "they" and how capable is this less-capable sensor shift based astro tracker?

I haven't shot astro beyond a single exercise in a landscape photography course, but my understanding is the farther a star is from the north star, the farther it moves across the frame, so sensor shift would be useful for stars close to the north star or short exposures, which would put a big emphasis on the less, rather than the capable.

 

[koenkooi]

Who is "they" and how capable is this less-capable sensor shift based astro tracker?

I haven't shot astro beyond a single exercise in a landscape photography course, but my understanding is the farther a star is from the north star, the farther it moves across the frame, so sensor shift would be useful for stars close to the north star or short exposures, which would put a big emphasis on the less, rather than the capable.

Pentax: https://milkywayphotographers.com/a...trotracer-guide-what-it-is-and-how-to-use-it/

 

[gruhl28]

Pentax: https://milkywayphotographers.com/a...trotracer-guide-what-it-is-and-how-to-use-it/

That looks pretty cool.

 

[MiJax]

Point is, once sensor shift is possible, you can do different things with it - IBIS, super resolution, AA filter, etc.

If one is shooting star trails, maybe the sensor can be shifted one pixel diagonally between each two shots to cover for dead / hot pixels. If shooting a static scene, it could shoot the scene three times, shifting the sensor to cover each pixel location with the three bands skipping the need for Bayer interpolation.

No dispute on your point, but its still a lot to go through for de-bayering. The only good reason so far is the ability to cancel it. However, who'd bet their paycheck that Canon will add the ability to cancel it out if they use the tech?

Using the pixel shifting for useful features makes sense, but this use just seems odd. The only other thing I can think of is... maybe its cheaper to manufacture without the AA so spending a couple of hundred thousand dollars to make that happen is worth it. IDK. From the outside looking in, just doesn't seem worthwhile... but hey, maybe we only have a portion of the plan.

 

[usern4cr]

Excerpt from patent's page 9:

"...when calculating the phase difference for focus detection, the influence of its spatial aliasing cannot be ignored, and in some cases, the obtained result of the focus detection may be largely erroneous. Therefore, conventionally, it is common to arrange the optical low-pass filter so that the response in the high frequency region higher than the Nyquist frequency is reduced..."

That is: the aliasing also causes problems to phase autofocus accuracy. So, contrary to some misconceptions, an AA filter is not an obstacle for the AF operation, but a "conventional" requisite to improve its actual performance. A bit of homogeneous blur doesn't mistake the AF system, which is even more fooled by the false detail. AA-free sensors are aliasing machines (nature shots have no less aliasing than its architecture counterparts: simply, our brain accepts as "sharpness" the part of random false detail).

Thanks! - This is the first time I've ever heard of the AA filter helping anything beyond moire reduction. I've always thought post processing (in camera or later) could replace an AA filter, but this may be the very reason that the AA filter is really needed, since AF needs to be done live and is crucial (to me) for sharp photos.