The end of Bayer?

John Sheehy said:

It's not even 100% true. The ouput image may sometimes appear as
sharp, but if you take a 6MP bayer image, upsample it 200% or 300%,
and then downsize it to 2268*1512 with the nearest neighbor
algorithm, then it looks similar to a Sigma/Foveon image. The
"sharpness" of the Sigma images is an artifact of the sampling
method, not sharpness composed of "image detail".

Click to expand...

I surely wish that someone would do all the experiments needed to settle this once for all.

It is not easy to do it - in fact it is very hard. It might even be imposible - due to the fact taht at the end it is subjective things that decide if an image looks good.

Roland

Laurence Matson said:

There is more than enough evidence around that many imagers are
reaching the physical limits of the pixel pitch. That is where
Moore's law does not apply.

Click to expand...

Moores law is a myth - just something that got true by repeating millions of times. If you look at diagrams over performance vs time you see no exponential curves - you see a big cloud that you can match to any curve you want - e.g. an exponential curve. Most evident is that during the last two years. If Moores law applied we would have 20 GHz processors now - but wo don't.

Laurence Matson said:

You cannot just keep cramming pixels,
which have a size limited by light waves, unto a limited space.
Folks can tell me all about their being sensor designers; if they
don't realize this, then I wish them lots of luck.

Click to expand...

Currently I guess that 1/10-1/5 of the light is detected by the sensor. So - there is a factor 5-10 there to gain. And you can always improve on the nosie level until you hit the photon nose level. After that you can do some clever tricks - using less time for exposing bright areas - taking a B&W image with some few extra color sensor .... There are lots you can do. All this could lead to more pixels per area. But - you are essentially right - there is a hard limit set by physics. But - we are not there yet.

On the other hand - there could be some technical break through that makes it possible to make giant sensors cheap. Polymer sensors maybe?

Roland

Yeah, the higher pixel densities would work well with this technique. I'm not sure that I agree that you'd get random patterns of stuff aligning with the random patterns of your sensor... I guess that in theory you'd get white-noise, so yes, there would be low-freq conmonents to the aliasing behavior, but it seems that white noise would be perceptually dominated by the high frequencies.

The behavior that seems the worst might be for a video camera based on the idea--you'd get sparkling where it couldn't deal with the high frequencies varying from frame to frame. But, I'm talking about fro still cameras.

Anyways, you make a good point about really only having a fixed random pattern. I think that we both agree that this could potentially be the best feature. We could keep people busy for the next five years debating wether or not they got a "good sample" of the random pattern!

Thanks for the comments,
Dave

Roland Karlsson said:

Oh yes - I am very happy with my Bayer cameras. Brilliant solution
really.

But there are problems. Where I live we have a large house with
lots of windows. This house I can only take photos of with my less
sharp lenses. If I try a sharp lens I get a very colorful result as
all blinds in the windows create aliasing. Lots of it. BTW - I have
Pentax *istD S.

Click to expand...

Try it with a D100.

Also, (serious suggestion) if RSE works with the Pentax, you might try that. The "image detail slider" controls the amount of Moire, from minimal to gross for test chart sorts of patterns, with the 300D. In images without Moire problems, cranking said slider does very nice things to contrast in fine detail.

I'd love a camera with interchangeable sensors. Then you can get the sensor you want: adequate AA, weak AA, or no AA.

--
David J. Littleboy
Tokyo, Japan

marrio said:

That is today. In the future we will need a 66Mp bayer sensor to
compete with the 22Mp Foveon type.

Click to expand...

If Foveon chips don't sell as well as Bayer sensors, we may never get to 22MP Foveon chips. Maybe they will forever be relegated to being a niche market due to not having been popular enough to sustain the R&D costs.

The thing about the X3 technology is that it produces 3* the data per sensor site as the Bayer sensors, so you will always have 1/3 of the resolution of the current technology. I don't know about you, but I'd rather have actual sensor site resolution and let the algorithms interpolate the color than measure the color with smaller, noisier sensors. And I like having 3FPS, and JPEG straight from the camera. And I like the build quality of Canon's EF lenses better than the Sigma lenses with EF mounts. I own 2 IS lenses that Sigma doesn't make competition for.

marrio said:

Maybe the 22Mp will be easier to make.

Click to expand...

The SD9/10 may be great 3.5MP DSLRs. But the 350D is better in terms of functionality, and produces 8MP of resolution with much better control of noise and fewer sensors (at more locations). When the technology is available to produce 22MP X3 sensors, will anyone care? By the time we get there, the Canon 650D will be available with 64MP at 3FPS giving clean ISO 3200 prints at 12x18 for $1000. Anyone who prints larger than that uses Medium Format (the 1Ds has broken into that market somewhat), and those have already gone past 25MP.

At any given point on the technology, Foveon sensors are 1/3 the resolution of their competition with the same file size. Eventually, the resolution won't matter all that much. At the moment, the Sigma/Foveon cameras are losing on the whole package. The framerate and AF system and IS and wide array of lenses and clean high ISO and variety of sensor sizes in the Canon system are all advantages.

-Mike
http://demosaic.blogspot.com

The Betterlight scan backs are range from 52 lp/mm to 125 lp/mm sampling fully measured color from lenses designed for 4x5 format. (Large format lenses tend to have lower peak MTF compared to the best 35mm lenses.) There is also some indication that they may overlap the sample positions slightly, but I can't find details on their site. (They mention "fractional pixels" but I don't see a definition.) There is also the possibility of jiggling the linear sampling array very slightly at each scan position, but I doubt the Betterlight camera does that.

(Joseph Wisniewski posted about the possibility of moving the sampling apparatus in a very small circle to accomplish low pass filtering in a discussion of Leica claiming to not need optical low pass filtering. Strikes me as a workable idea and that is the first place I'd heard of it. If one can get the mechanicals to work, it could even provide variable amounts of low pass filtering for different circumstances.)

The point here is that Betterlight is effectively already sampling above Nyquist for most circumstances in which they are used. I expect aliasing artifacts are very rare, though I have no personal experience with their products. They downside of very long capture times and inapplicability to moving subjects are well known.

-Z-

There is a nicely balanced article with samples at Luminous Landscape that shows some comp samples with 1DSM2 and LF Film. Again, they are clear to preface with the shortcomings, but it is interesting in the difference in approach. Fovean, Better Light, Canon's Patents, etc...it appears that there is more than one way to skin a sensor.

marrio said:

Maybe one more camera (1DIII) with 22Mp 9fps and Bayer (by this
time next year) before the RGB per pixel sensor. I think we are
still 4 years away form a 22Mp camera without low pass filters.

Márcio

Click to expand...

If it's not bayer, you don't need 22Mp to get the same information.

http://www.panavision.com/product_detail.php?maincat=1&cat=36&id=338&node=c0,c202,c203

we are probably near the end of Bayer.

--
http://www.andreabuso.com
http://www.andreabuso.com/select

You certanly have a point regarding the amount of data for the Foveon sensor. BTW, I had a Sigma lens but not anymore, now only Ls.

Márcio

they are probably using the same misleading pixel definition that Foveon is using. It is more likely a 4mp camera 4x4x4 rgb pixels. Bayer has proven to be the technology that has what it takes to go the distance. Why no 6mp foveon cameras yet?!?! Bayer surpassed the best foveon could offer almost immediately after the SD9 came out, and is now running circles around the SD10.

Sean Keegan said:

they are probably using the same misleading pixel definition that
Foveon is using. It is more likely a 4mp camera 4x4x4 rgb pixels.
Bayer has proven to be the technology that has what it takes to go
the distance. Why no 6mp foveon cameras yet?!?! Bayer surpassed the
best foveon could offer almost immediately after the SD9 came out,
and is now running circles around the SD10.

Click to expand...

Could be a 3-CCD system.

--
John

John Sheehy said:

Personally, I think that you have to have something wrong with your
perceptual mechanisms to think that an aliased SD9 image taken with
sharp optics is a realistic capture.

Click to expand...

Guilty as charged...

--

'Blessed is the man, who having nothing to say, abstains from giving wordy evidence of the fact.'

http://www.pbase.com/timothyo

Too much techno mumbo jumbo for this guy.

Just give me a camera that works great iso 100-800, 30" x 40" print size, and I'm happy. Who cares about the sensor?

It's not the technology, it's the result. Bayer or Foveon or who gives a darn what!

Steve Meier said:

Too much techno mumbo jumbo for this guy.

Just give me a camera that works great iso 100-800, 30" x 40" print
size, and I'm happy. Who cares about the sensor?

It's not the technology, it's the result. Bayer or Foveon or who
gives a darn what!

Click to expand...

Why are you participating in a thread that is of no interest to you?

--
John

Tim O'Connor said:

Guilty as charged...

Click to expand...

Do you see any problems with this image?:

http://www.pbase.com/dgross/image/31681259/original

--
John

Steve Meier said:

Just give me a camera that works great iso 100-800, 30" x 40" print
size, and I'm happy. Who cares about the sensor?

Click to expand...

30 X40 is a tall order at top quality. The prints can look good, but great is tough. And after all, we are really after the awesome, drop dead result. Even after looking at Medium and large frormat prints, the artist starts to think, "that could be better". That is really the distinction level we are discussing.

Again these are practical issues based on things like thermal, circuit and fixed noise, and the capture efficiencies of various photosensitive materals. These issues can all be managed by, as yet, undetermined tehcnological improvements.

I'm looking at this from the perspective of information theory and counting theory which will become the ultimatel limitations of these two technologies.

Each photon is the irreducible unit of colour information. If you filter photons out you are throwing away information.

--
Kent Dooley

(Some would argue that top end DSLRs are out-resolving lenses right now, but lets say at some point in the future an X megapixel sensor is able to out resolve ALL of Canons EF / EF-s lenses)

Bayer is a smart compromise between the amount of photosites available and the luminance resolution you can get. A Foveon sensor gives you much less luminance resolution all other things being equal, but the colour resolution is as good as the luminance resolution.

At some point, an X megapixel SLR will be as good as an X + 5 megapixel sensor, because the lenses cannot keep up. What's X? Lets say 24 to 30 megapixels, but that is just a guess.

You have gone as far as you can with Luminance resolution. So what do you do with the extra photosites that fabrication improvements give you? Improve the Colour resolution. A 100% then will have vibrant colour textures.

I wish to remind everyone that Kodak produced 1.3 factors since 1995's Nikon based DCS 460 and Canon Based DCS-1. This was followed by the DCS-560/660, and terminated in the 760. Canon got most of their "training" in digital camera operations from Kodak in the nineties, a smart move on their part.

I just ordered a 5D as based on all "goodness" factors of: pratical frame rate, full frame requirements, low noise, ideal pixel pitch, resolution, and Performance/Value for the DOLLAR. Most importantly, to service a new client which will spend 10 times the price of the new camera.

It will be my 32 nd. digital camera in 10 years, including a $50,000 Foveon (my most expensive). I didn't "lose" money on any of them as all of them enabled me to make more money for less cost than the equivalent "perpetual lease payment" called FILM in my Nikon/Hasselblad/Mamiya days. I don't miss optical printing or scanning, now that we have demonstrably superior inkjet and laser technologies to throw pixels on paper.

So if you break it down as "cost per click," superior image quality, and a systems approach to image SALES, digital wins by a long shot.

Without getting too "romantic" about more megapixels, blah blah blah, I made the most money with 1.5 to 2 megapixel output in the last 10 years of filmless photography.

One final note, all digital cameras SUCK at one applicaion or another. That's why I use 10 different ones depending on the application. They are simply a recording device that tells you about the quality of lighting, subject, and composition you put in the frame.

No one ever asked Pink Floyd what kind of tape recorder was used to create their masterpiece: Dark Side of the Moon.

I have not mastered photography. Like medicine, it is a PRACTICE.
--
Claude