Page 2 of 3 FirstFirst 123 LastLast
Results 21 to 40 of 41

Thread: Bayer vs Foveon Sensor

  1. #21

    Default Re: Bayer vs Foveon Sensor

    Resolution-wise there is little doubt about Foveon being able to capture more spatial color information. One thing still at doubt though is the accuracy of color reproduction, especially at low signal levels.

    Given the SNR perfomance of current Foveon chips, it is not too difficult to come up with an APS Bayer chip of 18-19MP to match the chrominance resolution of 14MP Foveon.

    Of course having to rely on Sigma to carry the chip to the mass market is difficult but of course we're only talking about the technology here.

  2. #22

    Default Re: Bayer vs Foveon Sensor

    Quote Originally Posted by CYRN View Post
    Anyway, I believe this article will address Bayer vs Foveon issue.
    Thanks for the very good article.

    And I quote " ... the SD14 really is fundamentally better technology".

    I think the Foveon is something to watch out for and not to be dismissed. If you do PS with very very fine selection you will realise the problem at edges due to Bayer's interpolation and I may go for Foveon just for that.

    And regarding "truth" I was really being ironical. For you need to know truth first to see truth: know to see and not see to know. If you are familiar with sampling theory you know about the Nyquist limit and so on.

    So you can always design a sensor - Bayer or otherwise - that gives you the resolution to pick the truth that you seeking: like detecting individual stars up to 10 million light years away. But you cannot see something you were never seeking, eg planets around those stars.

    But Bayer is unlikely for such applications for the interpolation really kills it: you never know what you see is the "truth" or just some algorithmic artefact. Most of them will be using multiple CCDs, and not just three, eg those in earth imaging satellites.

  3. #23
    Member
    Join Date
    Dec 2006
    Location
    Singapore, East Coast, Katong.
    Posts
    387

    Default Re: Bayer vs Foveon Sensor

    Quote Originally Posted by CYRN View Post
    If you are happy with 3MP handphone cams... by all means, use it as your main cam.
    Wasn't my post picked interpreted out of context??

    Quote Originally Posted by CYRN View Post
    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.
    now this is being quoted out of context.
    Last edited by XiaoMiaoWang; 2nd April 2007 at 09:11 PM.

  4. #24

    Default Re: Bayer vs Foveon Sensor

    don't be surprized if some other party comes up with 3-colour-in-each-sensor-location technology... the foveon way is only one way of doing it...

  5. #25
    Senior Member
    Join Date
    Sep 2004
    Location
    Cons digger.
    Posts
    3,924

    Default Re: Bayer vs Foveon Sensor

    Quote Originally Posted by CYRN View Post
    Mabbe this would be helpful in answering why your above statements are not really correct.
    that was exactly what I'm trying to put across. my qn on 4 sensor to 1 pixel on bayer CCD wasn't really a question.

    I was trying to demonstrate that pixel count in different sensor technology received different treatment. Even though the convetional bayer patterned sensors had interpolation done to generate a pixel it was accepted as 'true resolution', while SuperCCD's output was largely played down because the gurus said the output was 'interpolated'

    when foveon came along, they said it wasn't as good as conventional bayer sensors.
    “How fortunate for leaders that men do not think.” - Adolf Hitler

  6. #26

    Default Re: Bayer vs Foveon Sensor

    Quote Originally Posted by yanyewkay View Post
    that was exactly what I'm trying to put across. my qn on 4 sensor to 1 pixel on bayer CCD wasn't really a question.

    I was trying to demonstrate that pixel count in different sensor technology received different treatment. Even though the convetional bayer patterned sensors had interpolation done to generate a pixel it was accepted as 'true resolution', while SuperCCD's output was largely played down because the gurus said the output was 'interpolated'

    when foveon came along, they said it wasn't as good as conventional bayer sensors.

    45-degree Super CCD is "more interpolated" than 90-degree Bayer. The 90-degree Bayer has the highest resolution along 45-degree lines but Super CCD has the lowest resolution there. The argument was that human eyes are less sensitive to diagonal details, which is debatable.

    The problem with Foveon isn't the shear resolution but more with the color reproduction especially at high ISO (low signal levels).

  7. #27

    Default Re: Bayer vs Foveon Sensor

    Quote Originally Posted by grantyale View Post
    45-degree Super CCD is "more interpolated" than 90-degree Bayer. The 90-degree Bayer has the highest resolution along 45-degree lines but Super CCD has the lowest resolution there. The argument was that human eyes are less sensitive to diagonal details, which is debatable.

    The problem with Foveon isn't the shear resolution but more with the color reproduction especially at high ISO (low signal levels).

    I think you are comparing apples to oranges.Almost all the current sensors have to resort to algorithms when the ISO exceeds 200.So any fault with the colour reproduction may not be due to the sensor but with the algorithm used to do the interpolation.It will be interesting to use a Foveon sensor to capture the image and process it using Canon,Nikon or Sony technology.
    Last edited by Conundrumachinist; 4th April 2007 at 04:52 PM.

  8. #28

    Default Re: Bayer vs Foveon Sensor

    Quote Originally Posted by Conundrumachinist View Post
    I think you are comparing apples to oranges.Almost all the current sensors have to resort to algorithms when the ISO exceeds 200.So any fault with the colour reproduction may not be due to the sensor but with the algorithm used to do the interpolation.It will be interesting to use a Foveon sensor to capture the image and process it using Canon,Nikon or Sony technology.
    Not true, and noise reduction is a totally different issue. No matter what the incident level is, Bayer Pattern uses color filters to separate lights of different wavelengths. The only remaining difference is the signal to noise ratio. Also, interpolation algorithms do not change the statistical characteristics of the colors.

    In contrast, Foveon sensors separates colors by successive subtraction of signals obtained at different depth of the chip (red penetrates the furthest but get partial absorbtion along the way). This can lead to accumulation of noise (photon shot noise as well as dark current noise) in green and blue channel, coupled with weak red signal. Another issue is that the Foveon's absorbtion/response curves may deviate from the common RGB curves found in Bayer color filters.

    It is impossible to process Foveon Raw with others' algorithms as the working principles are different. Bayer pattern demosaicking works by "guessing" the missing color components at each pixel location based on information from the neighborhood; Foveon reconstructs color information at each pixel by subtracting signals at different layers.

  9. #29

    Default Re: Bayer vs Foveon Sensor

    Ok I'm not really an expert on this so bear with me.

    Quote Originally Posted by grantyale View Post
    Not true, and noise reduction is a totally different issue. No matter what the incident level is, Bayer Pattern uses color filters to separate lights of different wavelengths. The only remaining difference is the signal to noise ratio. Also, interpolation algorithms do not change the statistical characteristics of the colors.
    By increasing the ISO you introduce noise since above ISO 200 the signal from the sensor has to be amplified to simulate the higher ISO. I think this will be true for both Bayer and Foveon.But Bayer "throws away" the data of two colours for every pixel.This is where the interpolation comes in.So it interpolates and then amplifies.If you are taking a picture with various gradations of one of the primary colours (example a blue sky). The Foveon sensor should perform better.So I'm not too sure about your statement of interpolation algorithms not changing the statistical characteristics of the colours.

    Quote Originally Posted by grantyale View Post
    In contrast, Foveon sensors separates colors by successive subtraction of signals obtained at different depth of the chip (red penetrates the furthest but get partial absorbtion along the way). This can lead to accumulation of noise (photon shot noise as well as dark current noise) in green and blue channel, coupled with weak red signal. Another issue is that the Foveon's absorbtion/response curves may deviate from the common RGB curves found in Bayer color filters.
    On the point about the absorbtion this would be dependent on the thickness of each layer.I think it would be an easier problem to rectify.

    Quote Originally Posted by grantyale View Post
    It is impossible to process Foveon Raw with others' algorithms as the working principles are different. Bayer pattern demosaicking works by "guessing" the missing color components at each pixel location based on information from the neighborhood; Foveon reconstructs color information at each pixel by subtracting signals at different layers.
    Actually what I meant was the maturity of the technology.Foveon has been working on refining their sensor and they are a fairly young company.CCD sensors have been around longer so Canon, Nikon etc can concentrate more on the algorithmic aspects.If Foveon is able to put as much resources as them in processing their raw data their end product might be better.

  10. #30

    Default Re: Bayer vs Foveon Sensor

    The interpolation in a Bayer camera is done after amplification and digitization, but that's not the point. What I was trying to say was the channel separation is done with color filters as opposed to between-channel subtraction. This may fundamentally introduce some troubles when the signal level is low. By the way, I think the initial hype about improved saturation in SD9 has to do with the channel reconstruction (subtraction) with the raw data.

    I'll try to figure out the photon efficiency of the Foveon design to see the upper bound of SNR compared to Bayer, when I have time.

    In fact I like the idea of having all three colors at every pixel site. It's just that there are still some problems (both technical and marketing-wise) that prevent us from benifiting from the great idea.

  11. #31
    Senior Member
    Join Date
    Mar 2004
    Location
    Central West
    Posts
    1,952

    Default Re: Bayer vs Foveon Sensor

    There was some analysis done on the net a while ago based on the paper written by Foveon's chief scientist - the paper showed the equations and graphs for photon depth penetration vs wavelength. Can't remember the exact figures, but the analysis showed that for low light levels, a Foveon sensor looking at pure orange/red light will have 3x as much noise in non-red colours vs actual red signal, while a Bayer sensor will have no non-red noise, and around 0.3x noise vs signal in the red channel.

    The Foveon sensor tech isn't a total win solution. While it has advantages in some areas, it has some major disadvantages in other areas at it's current state (which may be why it isn't so widely adopted).

    You've got the noise issue.

    You've also got colour accuracy and metamerism issues. Because it's absorbing photons at different depths based on wavelength of the light, you have situations where "red" photons are absorbed by the higher layers and so register as "blue" or "green" photons. Also, the way it responds to colours is different from the human eye - for example you could have two objects that produce a spectral response in the eye where they look exactly the same colour. But if the spectral response of the sensor is not matched to the eye (as in the case of Foveon), those two objects will have a different colour in a Foveon image, whereas to the eye, they will look exactly the same.

    You've also got data transfer isssues. Most analysis says that a 3.4MP Foveon sensor is equivalent in resolution to about a 5-6MP Bayer sensor. But with the Foveon sensor, you need to transfer 10.2 MP (3.4 x 3) of data vs 5-6MP of data with the Bayer ie. 40-50% more data.

  12. #32

    Default Re: Bayer vs Foveon Sensor

    That's probably the propagation of noise I was talking about? Not sure though.

    gooseberry, can you please help to dig out the paper you are referring to? I'm looking for the figures.

  13. #33

    Default Re: Bayer vs Foveon Sensor

    Quote Originally Posted by grantyale View Post
    That's probably the propagation of noise I was talking about? Not sure though.

    gooseberry, can you please help to dig out the paper you are referring to? I'm looking for the figures.

    It could be " Foveon Technology and the Changing Landscape of Digital Camera Technology" found on their website http://www.foveon.com/article.php?a=74. There is data on penetration depth vs wavelength there.Sorry I can't post a direct link as the paper is in pdf format.

  14. #34

    Default Re: Bayer vs Foveon Sensor

    Thanks for the link. I went there a while ago but didn't notice they've put up all these papers.

  15. #35
    Senior Member
    Join Date
    Mar 2004
    Location
    Central West
    Posts
    1,952

    Default Re: Bayer vs Foveon Sensor

    Quote Originally Posted by grantyale View Post
    That's probably the propagation of noise I was talking about? Not sure though.

    gooseberry, can you please help to dig out the paper you are referring to? I'm looking for the figures.
    Yeah it's what you are referring to. As I remember someone writing once - the Foveon sensor efficiently converts photons to noise- heheheh....

    Can't remember the exact paper, but it was used as a reference in their patent application - I'll see if I can dig it up. Also, I think if you integrate under the graph of the absorbtion coefficient and penetration depth vs wavelength graphs in the paper Conundrumachinist pointed to above, you can get the probability that each wavelength will be absorbed at what depth.

  16. #36
    Senior Member
    Join Date
    Mar 2004
    Location
    Central West
    Posts
    1,952

    Default Re: Bayer vs Foveon Sensor

    Actually, speaking of multi-layer sensor design, Foveon isn't the only one. Fuji filed a few US patents in 2005 (they were originally filed in Japan in 2004). In these patents, Fuji described a multi-layer sensor design which seems to be much more promising than the Foveon tech. Instead of relying on the penetration depth response of silicon, it uses organic colour film to separate the colour layers, and so should have a response closer to film.

  17. #37
    Senior Member
    Join Date
    Sep 2004
    Location
    Cons digger.
    Posts
    3,924

    Default Re: Bayer vs Foveon Sensor

    Quote Originally Posted by gooseberry View Post
    Actually, speaking of multi-layer sensor design, Foveon isn't the only one. Fuji filed a few US patents in 2005 (they were originally filed in Japan in 2004). In these patents, Fuji described a multi-layer sensor design which seems to be much more promising than the Foveon tech. Instead of relying on the penetration depth response of silicon, it uses organic colour film to separate the colour layers, and so should have a response closer to film.
    i remember that too.. but so far no news...
    “How fortunate for leaders that men do not think.” - Adolf Hitler

  18. #38

    Default Re: Bayer vs Foveon Sensor

    Quote Originally Posted by gooseberry View Post
    Actually, speaking of multi-layer sensor design, Foveon isn't the only one. Fuji filed a few US patents in 2005 (they were originally filed in Japan in 2004). In these patents, Fuji described a multi-layer sensor design which seems to be much more promising than the Foveon tech. Instead of relying on the penetration depth response of silicon, it uses organic colour film to separate the colour layers, and so should have a response closer to film.
    Some patents are filed based on concepts without any actual development being done to "chope" the idea. I think we have to wait for a working prototype if and when it arrives.

  19. #39
    Senior Member
    Join Date
    Mar 2004
    Location
    Central West
    Posts
    1,952

    Default Re: Bayer vs Foveon Sensor

    Quote Originally Posted by yanyewkay View Post
    i remember that too.. but so far no news...
    Well there was speculation that since Fuji hasn't done much further with their current SuperCCD except bring it to their small digicams, that their engineers are working on this mulit-layer tech for a new product - they can't be doing nothing. Also there was speculation that Nikon were reluctant to go FF because they were waiting on the next generation sensor technology not based on Bayer - and considering the Fuji and Nikon relationship.....

  20. #40
    Senior Member
    Join Date
    Sep 2004
    Location
    Cons digger.
    Posts
    3,924

    Default Re: Bayer vs Foveon Sensor

    S5 just launched with a SuperCCD.. My guess is that there's at least another 2 years to wait.
    “How fortunate for leaders that men do not think.” - Adolf Hitler

Page 2 of 3 FirstFirst 123 LastLast

Bookmarks

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •