Apple patents a pentaprism

Allan Olesen said:

pgb said:

stevo23 said:

The novelty of it isn't in the splitting of colors but apparently how they do it. IE, "we found a new way to do this".

Click to expand...

That's true and it's also new for stills bar Foveon

Click to expand...

For still images, the concept of splitting different wavelengths out to several digital sensors is not new either. Minolta's first DSLR did exactly this.

But it was one green, one green and one blue/red sensor.

Click to expand...

I see, having green full resolution is good as most detail is derived from green. Sony had a still camera that pre-dated the Mavica, it was the size of two shoe boxes and to be used on a tripod, only $18k, I'm sure it was 3 chips.

So even less of a `new idea'.

The zoom optics in the patent would need to be top quality otherwise hue shifts would occur within
the frame. I can't see how all that will fit into a wafer thin phone. Maybe Apple is going to make small slr size cameras with a phone inside.

While discrete RGB capture is best, pro cine cameras are now all Bayer. I wonder if it's worth the trouble. I'd like to see raw converters have the option of interpreting Bayer as RGB albeit with the reduction in resolution just to see the difference. 50mpx sensors would lend themselves to this.

pgb said:

I'd like to see raw converters have the option of interpreting Bayer as RGB albeit with the reduction in resolution just to see the difference.

Click to expand...

You are suggesting to treat each 2x2 pixel group as a single RGB pixel?

I would assume that it would create some rather serious aliasing artifacts on subjects with fine colour detail because the "centre of gravity" of the R, G and B information would be offset from the centre of the pixel.

You could insted use 3x3 pixel groups to form a single RGB pixel. This would have the centre of gravity for all 3 colour channels aligned with the centre of the pixel. But each of those pixels would not have the same distribution of R, G and B subpixels.

Count me among those who doesn't see how this would fit in a phone. Apple may be trying to break into the camera market.

edu T said:

...but, unfortunately, they happen to already have launched a product line called, by sheer coincidence, NEWTON.

Anyway, I think it's worth noting that said prisms are NOT the part labeled deflector 20 on the drawing the OP posted, but rather the light splitter 2 assembly on the bottom of the optical stack, which is the "cube of triangular prisms" itself:

It's about a Newtonian way of splitting light...

Click to expand...

Actually, it's not. If anything, it's Zhenderian. But I wouldn't even call it that.

It's a technique for splitting light into three parts via partially silvered beam splitters. The color separation is still done by filters. The filters can't be dichroic beam splitters, they pretty much have to be on the sensors, and they can't be dichroic for reasons that should be glaringly obvious.

Problem is, it doesn't actually work very well, and I'll explain why in a bit...

--
Rahon Klavanian 1912-2008.
Armenian genocide survivor, amazing cook, scrabble master, and loving grandmother. You will be missed.
Ciao! Joseph
www.swissarmyfork.com

Allan Olesen said:

pgb said:

I'd like to see raw converters have the option of interpreting Bayer as RGB albeit with the reduction in resolution just to see the difference.

Click to expand...

You are suggesting to treat each 2x2 pixel group as a single RGB pixel?

I would assume that it would create some rather serious aliasing artifacts on subjects with fine colour detail because the "centre of gravity" of the R, G and B information would be offset from the centre of the pixel.

You could insted use 3x3 pixel groups to form a single RGB pixel. This would have the centre of gravity for all 3 colour channels aligned with the centre of the pixel. But each of those pixels would not have the same distribution of R, G and B subpixels.

Click to expand...

I was thinking just 3 pixels. Would it have more aliasing than Bayer with an AA filter?, I'd have to think about that. The light entering a 2x2 would still overlap the next pixel group with Bayer.

That may be why it hasn't been done but the reduced resolution is probably unacceptable or it's
not worth the trouble. We have an RGB 444 video monitor at work which looks more photographic
with sharper more defined colour but's that comparing it to 422 YUV which is different to Bayer.

Foveon does seem to produce a different look, so their must be advantages to RGB.

One of the first Sony cine cameras was single sensor RGB. The sales person was quick to point out,
`no bayer' but that's history now. So it can be done - in HD.

Ah, the old double cube beam splitter. I think every newbie optical engineer has been sucked into it's promise of short optical paths at one time or another.

The only problem is that it's essentially useless.

Everyone's familiar (or can easily become familiar) with the dichroic beam splitters used in old-fashioned "3CCD" video cameras. They offered near 100% filter efficiency, at the cost of weight, bulk, and (most important) optical path length about 3x the sensor height. This last forced the lenses to have enormous compromises in their optical designs. (think "infinity optics").

So, every now and then, someone thinks this:

• What if we didn't just split the cube one one diagonal and send light in two directions? What if we cut and silvered two diagonals and sent light in three directions. That way, we wouldn't have to have two cubes and double the optical path length. We could do it all in one cube.

Then they either think about it for a while, simulate it, or have the lens makers make a prototype. And that's when they learn:

1. It doesn't split light three ways, it splits light four ways, and the fourth way is right back into the rear element of the lens. Not just a little light, like a poorly coated sensor IR filter, but 25% of the light. See the last and second from last elements. That's four surfaces nearly perpendicular to the light path. You though the ghosts in Sony's silly "transluscent" mirror SLRs were something? This is a whole different league.
2. At small effective apertures, you see the intersection of the two cuts stretching across the center of the image. It also causes flare.
3. They're plain old beamsplitters, not dichroic, so each sensor gets 25% of the light, and then it gets filtered. It's actually less efficient than a Bayer pattern sensor, which sends 25% of the light to a red filter, 25% to a blue filter, and 50% to a green filter.

So, you compromise the optical design to accommodate a splitter, dramatically increase flare, increase cost and complexity (dramatically, because we have three sensors mounted in different planes) and decrease efficiency compared to what we have now.

It's a win-win.

(Provided you redefine "win")

Hithertoo said:

Hithertoo said:

Getting the perfect photo of one’s lunch or night out with friends may be about to get a lot easier. As AppleInsider reported, Apple has been awarded a patent for a new type of sensor for smartphone cameras that could dramatically improve how images are captured by its iPhones. The patent uses the same technology that’s found on this seminal prog-rock album cover: prisms.

According to the patent, the digital cameras found in most smartphones have a single light sensor. That has a color filter split between red, blue and green colors, laid in a grid on top. Once an image is captured by the light sensor, it has to reconstruct the image based on the samples of color from the filter with a process called demosaicing. This tends to lead to images that are blurrier than real life, as the light sensor can only pick up colors where they are in a grid.

Apple’s new idea: Using a cube of triangular prisms, light is split out into green, red and blue streams, which are each picked up by an independent sensor. This would mean images would be truer to life in color and resolution. It could also mean that photos taken in low-light situations (indoors, at night, etc.) would actually be clearer.

Click to expand...

http://qz.com/368989/apples-new-digital-camera-patent-uses-prisms-to-create-more-lifelike-photos/

Click to expand...

--
Rahon Klavanian 1912-2008.
Armenian genocide survivor, amazing cook, scrabble master, and loving grandmother. You will be missed.
Ciao! Joseph
www.swissarmyfork.com