NEW DP QUATTRO is coming!!!!!!!!!!!!!

Started Feb 9, 2014 | Discussions thread
Raist3d Forum Pro • Posts: 36,758
Re: NEW DP QUATTRO is coming!!!!!!!!!!!!!

DMillier wrote:

This is still a bit vague for me! trouble is I'm a bit thick and need this spelled out

Let me explain my own internal rationalisation so you can understand my thought processes (and hopefully spot the error).

Let's imagine we are photographing a chess board type thing with the camera neatly arranged so one square on the board exactly fits one pixel in the final picture. That means each square is aligned with each 3 photosite triplet stack. Let's also imagine that this is a very odd chessboard in that all the squares are white except for one in the middle of the board. Let's also imagine that the black square is the one we are interested in.

So with the traditional X3 sensor (and let's pretend it's a RGB sensor), the black square (say, RGB 5,5,5) will record as RGB 5,5,5 and all the surrounding white squares (say RGB 250,250,250) will record accurately as 250, 250, 250.

With the quattro things are somewhat different.

The top layer will accurately measure the black square's blue value and accurately measure the white squares' blue values. Good.

But the g and r layers can't see the little black and white squares because the photosites are 4 times bigger. So what they will see is the black square + bits of lots of surrounding white squares.

And the only thing they can do with information is average it all together to produce and average green value from blending together everything in the (large field of view) of the green photosite and an average red level from the field of view of the large red phototsite. Those average levels g and r values will then be fed back to all 4 of the top layer pixels

So the quattro top layer will see blue value of pixel 1, blue value of pixel 2, blue level of pixel 3, blue level of pixel 4

the green layer will see a bit of green from square 1 + a bit of white from the next square + a bit of white from the next square + a bit of white from the next square.

The same from the red layer.

So, in effect, only the blue channel sees one square at each spatial location, the green and red layers will see a mix of neighbouring squares that is the same for all 4 of the small blue layer pixels.

This means that at every spatial location the colour is wrong because it has an accurate value for the blue value and a fudged average of the black square and several of its white neighbours for the green and red layers.

In a way, the way you put it, you are right.  If the chess squares aligned in size and spacing with the red/green square then it would be sampled correctly, so I wouldn't say always.

So you are right vs a true X3 Foveon, there is going to be some color error. I think the best way to see it is that color resolution (full color resolution) lowers while the sensor still picks up high resolution for the blue and because we are much more sensitive to luminance, for many cases, the approximation will probably be pretty reasonable using the blue layer as guidance for the luminance.

Keep in mind though, that yes, there's a bit of wrong here, but vs Bayer, at least assuming the noise isn't too bad this time, the X3 Quattro would be able to still measure spatially all blues, and the greens are reds are just lower resolution, not a grid of holes which is the case with Bayer and Xtrans.

So I think this will help in keeping some degree of accuracy better than what Bayer does, and having a full layer that is full resolution will surely help to hint the luminance better than the green with holes of Bayer or Xtrans, even when both have more greens than their reds and blues (the X3 Quattro still has full resolution blue, Xtrans and Bayer never have full resolution of one color).

So in short- what you are describing is correct. It will require some guess strategies but it should still be better than Bayer (potential noise issues aside) because you are still capturing more data anyway and better spatially aligned.

-- hide signature --
-- hide signature --

Raist3d/Ricardo (Photographer, software dev.)- I photograph black cats in coal mines at night...
“The further a society drifts from truth the more it will hate those who speak it.” - George Orwell

Post (hide subjects) Posted by
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow