Bayer vs Foveon Sensor


Status
Not open for further replies.

espion

Deregistered
Aug 25, 2005
1,524
0
36
A Mike Chaney have done some analysis and these are his conclusions:

* Bayer sensors have a very significantly reduced resolution when resolving detail comprised of mainly red/blue primaries, such as a red sports car with black pin stripes, a blue sweater with red lettering, red soda can with black lettering, etc. In these cases, resolution of the Bayer sensor is reduced to less than 1/4 of its "image" resolution! Black and white details will show the highest resolving power on a Bayer sensor, while saturated color detail will vary greatly. A Foveon sensor is much more consistent, resolving near the full resolution of the images for every color combination.

* Bayer sensors will produce images that are softer and less detailed due to the "smoothing" needed to eliminate artifacts and color distortions.

* Bayer sensors tend to omit chrominance (color) information when sampling high frequency detail. If you look at a picture of a tree that has many small branches with a brick wall behind it for example, you will see that many of the smaller branches "morph" into the color of the bricks in the background. This is because the branches are not wide enough to cover the multiple pixels needed to derive accurate color information on a Bayer sensor. Full color sensors completely eliminate this problem.

* Bayer sensors tend to produce color moire on high frequency detail like the cuff of the jeans below (more visible in center image). (see link above for images)

 

Fevon sensors have relatively bad noise control. Details are also lost when shooting at higher ISOs. IIRC, the NR algorithm isn't as good too.

Fevon sensors seems to have stagnanted in their offering of high MP sensors (have to discount the 3 layers as it does not add to resolution), the MP race in bayer sensors are now already out-resolving the lenses themselves.
 

have to admit the AA filter dose dramatically soften the images, it dose get rid of the horrible moire though
 

have to admit the AA filter dose dramatically soften the images, it dose get rid of the horrible moire though

:bsmilie: Ironically this may partly contribute to the allure of the 1D and 1Ds Mark I. They have weak AA filters, thus leading to sharper images out of the camera. All this is at the expense of getting occasional moire however, as you mentioned.

Foveon sensor technology is very interesting and the recent SD14 is indeed attractive. But I worry about lens compatibility and the future viability of such a system.
 

Fevon sensors have relatively bad noise control. Details are also lost when shooting at higher ISOs. IIRC, the NR algorithm isn't as good too.

Fevon sensors seems to have stagnanted in their offering of high MP sensors
(have to discount the 3 layers as it does not add to resolution),

huh? what do u mean? discount the 3 layers????


the MP race in bayer sensors are now already out-resolving the lenses themselves.
 

huh? what do u mean? discount the 3 layers????

dpreview-SD14 specs said:
Photodetectors 14.1 million (2652 x 1768 x 3)
Effective pixel locations 4.6 million (2640 x 1760)

Image sizes
• 4608 x 3072 (Super High) - Interpolated
• 2640 x 1760 (High)
• 1776 x 1184 (Medium)
• 1296 x 864 (Low)

Fevon have only 4MP (real) rez. Good enough for most. But it sort of defeats the above reasoning that Bayer have 1/4 resolution in specific situations when Bayer sensors are now 10MP and reaching 12MP soon.

Also, a point to note is that the below article was done in 2002. Technology have already left it behind. ;)
 

Foveons maintain near total colour fidelity; but stuck at 4mp of real resolution, it does not lend itself well to landscapes.

Double the MPs, then we're talking.
 

Fevon have only 4MP (real) rez. Good enough for most. But it sort of defeats the above reasoning that Bayer have 1/4 resolution in specific situations when Bayer sensors are now 10MP and reaching 12MP soon.

Also, a point to note is that the below article was done in 2002. Technology have already left it behind. ;)

not really understand what your saying
but then again... not a serious photographer..
though imaging is my living.

Feveon has 4 Megapixel, each pixel yields three colors.
From information theory, if each color has 10 bits resolution then each pixel yields 10x3 bits information.
For a Bayer sensor of 4 Megapixel, each pixel yields ONE color.
Hence similarly each pixel only yields 10 bits of information.
To get 3 colors per pixel, surrounding pixels are made sensitivity to each one of RGB using Bayer pattern, and then interpolated. Hence for each pixel, one color is measured and two others are interpolated or for each pixel the sensor is color blind to 2 of 3 primary colors.
IOW, a bayer type sensor can be in your terminology be termed as 12M (fake) which imho wld then be similar to foveon 4M (real )

Now if there were 4 primary colors instead of 3 then, there would be no need for bayer pattern but jus simple 4 color equally distributed pattern. This would make comparison of the 2 sensors much simpler.
However for 3 primary pattern on monochrome sensor, the bayer pattern becomes necessary and Green being the color repeated. Comparison becomes difficult as there is now no common denominator.

Without proof, inkling is that foveon will be superior assuming both are perfect sensor without noise problem as it gives the sensor values at the correct position.

discuss la.
 

discuss la.

Not really prove, but just some thoughts...

As a photographer, my interest is only 4 things.

1. Absolute resolution (determines how large I can present my photo)

2. Noise

3. Dynamic Range (needless to say more is better in most cases)

4. Color fidelity and accuracy (personally I feel that limiting factors are more on the output rather than input - monitors and printers are the constraints)

In bayer, the interpolation is only for colors, the drawback is that in certain color combination condition the interpolation alogrithm is not ideal, thus you lose resolution in such condition and it's only for a portion of the whole photo. ie the case of moire of a bridal gown dosen't affect the resolution of the bride's face.

Anyway, my personal opinion is that it's possible to identify such conditions and have another type of alogrithm to process such interpolation while maintaining resolution. Maybe it's still a dreamware as identification of such conditions are very tricky with today's techniques and might give a lot of "false-positive".

So, for now my stand is still for more native resolution directly affected by above factors 1, 2 and 3, as colors can be tweaked in post process anyway.
 

As a photographer, my interest is only 4 things ...
But you stay with digital nonetheless and not go back to film for its greater "truth"?
 

But you stay with digital nonetheless and not go back to film for its greater "truth"?

The thing with flim is that they have existed for "ages" and when digital cams were introduced, flims were used as benchmarks.

However, that just should be it...benchmarks... and nothing else. Don't forget, we are using consumer grade devices, non of which can be compared to the Digital cams of Hubble's and Mars Rovers... that's where the REAL digital performance lies... only question is that are you willing to pay. (BTW, I believe the above examples uses Bayer.. ;) )

If people keep comparing "truths" then we should all stick to creating arts ala Da Vinci way... paintings... where brush strokes, paint mixture and paper types are the "real truths".
 

Foveon is good for pixel-peeping, when Sigma is about the only one making cameras with it. I am not sure whether the noise characteristics has to do with the fact that Foveon is more densely packed. The color reproduction from Foveon doesn't seem to be very consistent with the trichormatic conventions at different incident levels. The often boasted "more saturated color" makes little sense because of the (inevitable) use of between-channel subtractions in raw conversion of Foveon.

The lack of AA filter in Sigma cameras isn't technically justified as well. But then it's more of a Sigma issue I suppose?

In terms of color fidelity, I think digital is better than film. But then, it doesn't have to be digital, the signal comes from near-linear analog devices anyway.
 

But you stay with digital nonetheless and not go back to film for its greater "truth"?

WHAT TRUTH? Either I am blind, my LCD got color banding, or I cannot see the word "truth" in the previous posts.

And if shoot film, it means I've attained the ultimate nirvana in attaining 'truth' in photography?

What nonsense.

What about the works of brassai, man ray, hockney, weston, and so many others who have pursused abstraction in their photographs? So is abstraction the 'truth'? far from it. Go back and read up lah....
 

Not really prove, but just some thoughts...

As a photographer, my interest is only 4 things.

1. Absolute resolution (determines how large I can present my photo)

2. Noise

3. Dynamic Range (needless to say more is better in most cases)

4. Color fidelity and accuracy (personally I feel that limiting factors are more on the output rather than input - monitors and printers are the constraints)

U mean image sensor tester?
 

Well I never said or implied that...

Especially when you make a blanket statement by saying that all photographers should or must give a damn about image sensors...

nether did I make any blanket statements... you quoted me out of the context from my reply to richardg. :nono:

My last 2 para that you have left out of your quote also reinforced that's it's my personal opinion.

Anyway, by the manner you've replied my last post, it does imply that some of the 4 factors does have some influence on yourself. Cuz you've indirectly agreed that 3MP handphones cams is not suitable for your main cam. ;)
 

with regards to pixel count and 'absolute resolution'. Bayer patterned sensors generally require 1 sensor for each colour usually arrange in RGGB, so 4 sensors generate 1 coloured pixel.
http://www.foveon.net/files/PixelLocation.gif
in case those who doesn't know :eek: yes there are 2 times Green pixels. So 50% of your sensor is covered with green sensitive pixels

Take a camera in the market, sold as 6 mega-pixels. Output file 3000x2008 pixels (6millions pixels), so how does 6 million sensors translate to 6 million coloured pixels? shouldn't 4 sensors generate 1 coloured pixel resulting only in 1.5 million coloured pixels?

Applying the same theory, we would have a 12megapixel camera from a Fuji SuperCCD camera, however, it was largely played down as 'interpolation' as the 12megapixels didn't contribute to 1 pixel for each sensor. :think:

So, if going for pixel count, I think Fuji wins hands down. For the "true" absolute pixel count, Foveon sensors speak the truth.
 

with regards to pixel count and 'absolute resolution'. Bayer patterned sensors generally require 1 sensor for each colour usually arrange in RGGB, so 4 sensors generate 1 coloured pixel.
http://www.foveon.net/files/PixelLocation.gif
in case those who doesn't know :eek: yes there are 2 times Green pixels. So 50% of your sensor is covered with green sensitive pixels

Take a camera in the market, sold as 6 mega-pixels. Output file 3000x2008 pixels (6millions pixels), so how does 6 million sensors translate to 6 million coloured pixels? shouldn't 4 sensors generate 1 coloured pixel resulting only in 1.5 million coloured pixels?

Applying the same theory, we would have a 12megapixel camera from a Fuji SuperCCD camera, however, it was largely played down as 'interpolation' as the 12megapixels didn't contribute to 1 pixel for each sensor. :think:

So, if going for pixel count, I think Fuji wins hands down. For the "true" absolute pixel count, Foveon sensors speak the truth.

Mabbe this would be helpful in answering why your above statements are not really correct.
 

Anyway, I believe this article will address Bayer vs Foveon issue.
 

Status
Not open for further replies.