Go Back   Steve's Digicams Forums > Digicam Help > General Discussion

Reply
 
Thread Tools Search this Thread
Old Feb 27, 2006, 5:19 PM   #1
Junior Member
 
Join Date: Sep 2004
Posts: 8
Default

Foveon released sensors some time back that image all three primary colours per pixel. They foolishly marketed a 3mp sensor as 10mp though. What's happened since then? They're in a few cameras, but nothing mainstream. Canon and Nikon Digital SLRs would really be great if they got rid of the bayer filtering or such means of synthesizing the colour. Thanks,

David
redcliffe is offline   Reply With Quote
Sponsored Links
Old Feb 27, 2006, 9:57 PM   #2
Senior Member
 
Join Date: Dec 2002
Posts: 5,803
Default

You're not the only one wishing this sensor would go some where.
It is a great idea and the images it can produce seem very good. It has a variety of advantages over the CCD/CMOS sensors.

My guess (just a wild guess!) is that it could scale up. They couldn't make the photosites (the things on the sensor that sense the light) small enough to get the high resolution that people demanded.

I wish Canon (since I'm a Canon user) would buy them out and push it to completion. But I'd have thought it would have happened by now if it ever would.

Eric
eric s is offline   Reply With Quote
Old Feb 28, 2006, 10:21 AM   #3
Senior Member
 
Join Date: Jul 2003
Posts: 3,396
Default

Too true, I seem to remember way back when it first came out there was talk of Canon and other using it. It never seemed to pan out and everyone went their own way. The only available fovon based camera I know of is the aging sigma sd10.

Maybe they will have something new up their sleeves soon, as the fovon did seem to be a good technology.

PeterP is offline   Reply With Quote
Old Feb 28, 2006, 11:07 AM   #4
Senior Member
 
Join Date: Oct 2004
Posts: 448
Default

I personally think that the Foveon sensor is a clear failure and that it will be discontinued within the next years. The claim of Sigma that their camera has 10 MP instead of 3.3 is simply a lie. Actually a 3.3 MP Foveon sensor has hardly more real pixels than an ordinary RGB sensor based on a Bayer filter grid. An RGB sensor records only one color per pixel. 50% of the pixels are green, 25% are blue and the remaining 25% are red. The remaining colors, i.e. two colors per pixels, have to be interpolated. This process is called demosaicing. In camera demosaicing must be simple, whence quite a bit of resolution gets lost. I would guess that the real resoution of an RGB sensor with in camera processing is only 40-50% of the claimed resolution. However, raw image processing on the PC allows to use very sophisticated demosaicing algorithms, which can squeeze out more detail from the sensor output. I would estimate the real resolution obtained by good raw processing software at about 60% and we certainly can get to 70%. Thus if you shoot raw you have a real resolution of about 60% of the nominal resolution. The Foveon sensor records 3 values per pixel, but that doesn't mean that its real resolution is 100%. I would guess it is 80% or even lower, because the colors are not well separated. It too has to apply some kind of demosaicing. Actually raw processing of the output of the Foveon sensor is more complicated than that of RGB sensors. In reality the real resolution isn't much better than that of RGB sensors. But because it is more complex, most of the recent progress concerning RGB sensors, cannot be transferred easily to the Foveon sensor. Hence, while the ordinary sensors get better and better, there is no progress with the Foveon sensors and that's the death nail for this sensor.
kassandro is offline   Reply With Quote
Old Feb 28, 2006, 12:22 PM   #5
DBB
Senior Member
 
Join Date: Jan 2004
Posts: 2,483
Default

kassandro wrote:
Quote:
I personally think that the Foveon sensor is a clear failure and that it will be discontinued within the next years. The claim of Sigma that their camera has 10 MP instead of 3.3 is simply a lie. Actually a 3.3 MP Foveon sensor has hardly more real pixels than an ordinary RGB sensor based on a Bayer filter grid. An RGB sensor records only one color per pixel. 50% of the pixels are green, 25% are blue and the remaining 25% are red. The remaining colors, i.e. two colors per pixels, have to be interpolated. This process is called demosaicing. In camera demosaicing must be simple, whence quite a bit of resolution gets lost. I would guess that the real resoution of an RGB sensor with in camera processing is only 40-50% of the claimed resolution. However, raw image processing on the PC allows to use very sophisticated demosaicing algorithms, which can squeeze out more detail from the sensor output. I would estimate the real resolution obtained by good raw processing software at about 60% and we certainly can get to 70%. Thus if you shoot raw you have a real resolution of about 60% of the nominal resolution. The Foveon sensor records 3 values per pixel, but that doesn't mean that its real resolution is 100%. I would guess it is 80% or even lower, because the colors are not well separated. It too has to apply some kind of demosaicing. Actually raw processing of the output of the Foveon sensor is more complicated than that of RGB sensors. In reality the real resolution isn't much better than that of RGB sensors. But because it is more complex, most of the recent progress concerning RGB sensors, cannot be transferred easily to the Foveon sensor. Hence, while the ordinary sensors get better and better, there is no progress with the Foveon sensors and that's the death nail for this sensor.
Sigma has stated they will release a new camera in the near future, based on the Foveon sensor.

For all the problems with the Foveon sensor, I believe that it is in fact a clear success.

Nor do I hold their claim to be 10 megs a lie.

They were faced with the problem of comparing their sensor with the bayer based sensors. This is an apples and oranges comparison. It can't be done. If they would have stated only 3 plus megs that would have been equally a "lie."

Moreoever, to say that they require the same amount of interpolation to achieve an image is simply not the case.

People with far more technical expertise then I, and with no axe to grind, state emphatically that the 3 plus Foveon is the equivalent of a 6 or greater bayer sensor.

The problems with the Sigma cameras were minor in a technical sense. If there had been money behind the research, I'm sure they would be worked out. Indeed, perhaps Sigma HAS worked them out.

There were occasional color problems in skies and a tendancy for greater aliasing of images. These were more then compensated for by the incredible clarity of the images, which allowed far more interpolation up.

Foveon images clearly print better then Bayer images. I can say this with some assurance since I make a living printing 16 x 20's and printed out samples of both from an independant source, as well as my own images.

What the latest Sigma release will be is hard to say, since this company always plays it close.

Dave
DBB is offline   Reply With Quote
Old Feb 28, 2006, 3:55 PM   #6
Senior Member
 
Join Date: Oct 2004
Posts: 448
Default

Look at Dcraw.c. It handles all kind of sensors. There are at least three Bayer demosaicing algorithms in dcraw.c and all three together are smaller than the one used for the Foveon sensor. I haven't looked at the details of the Foveon algorithms, but the Foveon statement, that their statements simply captures all three basic colors at one pixel is bluntly false. Actually proccesing of the Foveon sensor data is so complicated that it can't be done in camera. You can't compare the big Foveon sensor (crop factor 1.7) with the tiny sensor of a P&S camera though. You have to compare it with a APS-C size sensor of a DSLR like my KM 7D (crop factor 1.52) and I'm quite confident that raw images of my KM 7D processed with the free RSE are more than competitive with those of the Sigma SD9 although it has only a whimpy 6 MP compared with the Sigma's 10 MP.
kassandro is offline   Reply With Quote
Old Feb 28, 2006, 5:43 PM   #7
DBB
Senior Member
 
Join Date: Jan 2004
Posts: 2,483
Default

kassandro wrote:
Quote:
Look at Dcraw.c. It handles all kind of sensors. There are at least three Bayer demosaicing algorithms in dcraw.c and all three together are smaller than the one used for the Foveon sensor. I haven't looked at the details of the Foveon algorithms, but the Foveon statement, that their statements simply captures all three basic colors at one pixel is bluntly false. Actually proccesing of the Foveon sensor data is so complicated that it can't be done in camera. You can't compare the big Foveon sensor (crop factor 1.7) with the tiny sensor of a P&S camera though. You have to compare it with a APS-C size sensor of a DSLR like my KM 7D (crop factor 1.52) and I'm quite confident that raw images of my KM 7D processed with the free RSE are more than competitive with those of the Sigma SD9 although it has only a whimpy 6 MP compared with the Sigma's 10 MP.
LOL

First of course they don't make the claim that each sensor captures all three colors. The Foveon sensor, uses three sensors and creates a pixel. Since this pixel is a true color reading, there is less interpolations involved.

And that claim is true.

Your point is that it's a lie to claim 10 megs. Wouldn't it also be true that it's a lie that your Bayer sensor is 6 Megs, because those sensors actually interpolate their pixels?

"The cone-shaped cells inside our eyes are sensitive to red, green, and blue—the "primary colors". We perceive all other colors as combinations of these primary colors. In conventional photography, the red, green, and blue components of light expose the corresponding chemical layers of color film. The new Foveon sensors are based on the same principle, and have three sensor layers that measure the primary colors, as shown in this diagram. Combining these color layers results in a digital image, basically a mosaic of square tiles or "pixels" of uniform color which are so tiny that it appears uniform and smooth."

"All other digital camera sensors only measure the brightness of each pixel. As shown in this diagram, a "color filter array" is positioned on top of the sensor to capture the red, green, and blue components of light falling onto it. As a result, each pixel measures only one primary color, while the other two colors are "estimated" based on the surrounding pixels via software. These approximations reduce image sharpness, which is not the case with Foveon sensors. However, as the number of pixels in current sensors increases, the sharpness reduction becomes less visible. Also, the technology is in a more mature stage and many refinements have been made to increase image quality."

http://www.dpreview.com/learn/?/Glos...sensors_01.htm

Color Accuracy

"Conventional sensors using a color filter array have only one photodiode per pixel location and will display some color inaccuracies around the edges because the missing pixels in each color channel are estimated based on demosaicing algorithms. Increasing the number of pixel locations on the sensor will reduce the visibility of these artifacts. Foveon sensors have three photodetectors per pixel location and create therefore a higher color accuracy by eliminating the demosaicing artifacts. Unfortunately their sensitivities are currently lower than conventional sensors and the technology is only available in a few cameras."

http://www.dpreview.com/learn/?/key=pixel_quality

The above stolen from DP Review.
DBB is offline   Reply With Quote
Old Feb 28, 2006, 6:22 PM   #8
Senior Member
 
Join Date: Oct 2004
Posts: 448
Default

I completely agree with all what you say about Bayer type sensors. I never claimed that a 6 MP sensor really yields a 6 MP resolution. It doesn't do that and will never do that. You are right, a Bayer sensor measures only the luminance of the incoming light. The colors are measured with a filter grid, the Bayer mosaic, above the electronic layer. The advantage of this design is its simplicity. The semiconductor engineers have only to worry about greyscales while the optical engineers deal with the filter grid, micro lenses, antialiasing etc.. In the Foveon sensor these things are no more separated and that makes the whole matter rather messy. That makes progress much harder.
The statements and nice pictures about the Foveon are bluntly false. If they were true, then no data processing at all would be required. In reality, the Foveon sensor data require more processing than the Bayer sensor data. Consequently, the Sigma SD9 only supports a raw format. At the time when the Sigma SD9 came to market memory cards were much more expensive than today. They would have loved to deliver small jpeg images but they couldn't. When I saw that this camera only supports raw, I quickly recognised that the beautiful pictures about the Foveon sensor are pure fraud.
kassandro is offline   Reply With Quote
Old Mar 1, 2006, 8:03 AM   #9
Senior Member
 
Join Date: Aug 2002
Posts: 851
Default

And all of this 'my sensor is better than your sensor' is meaningless, when Foveon/Sigma cannot produce a camera capable of competing with the current offerings from Nikon, Canon etc.

If Foveon/Sigma want to prove their worth, AND stay in the game, they need to produce a new camera - now. Otherwise they are just an also ran.
amazingthailand is offline   Reply With Quote
Old Mar 1, 2006, 10:08 AM   #10
Senior Member
 
Join Date: Dec 2002
Posts: 5,803
Default

I'm with amazingthailand here. Foveon has to step up and make a new sensor capable of competing with where the market has gone. I'd like them to succeed. I really would. I know some people who really like the output of that sensor, and they know what they are talking about.

As a photographer, I don't care if the cameras that use a Foveon sensor can't produce a jpg. I don't care if the sensor is difficult to engineer compared to other sensors. I don't care if it takes a lot of computing power to produce the image from the RAW. (As an engineer by training, I do care about those things... but only at an intelectual level.)

I care about quality of output image and usability of the camera. And the quality of the image from a Foveon sensor based camera is quite good, from all I've seen. This is why I want them to succeed. Competition in the market is a very good thing, and JFET seems to have failed in its current incarnation.

So either Nikon needs to improve its JFET design, Foveon needs to improve its design or something new needs to appear. That will spur new growth in the market of sensors and everyone will benefit.

Eric
eric s is offline   Reply With Quote
 
Reply


Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are Off
Refbacks are Off



All times are GMT -5. The time now is 10:52 AM.