Foveon X3 sensor, entropy, and Bayer sensors

Discussion in 'Digital Photography' started by William Wallace, Jan 18, 2004.

  1. Okay, so I have been thinking about this. A Foveon sensor does
    capture more information. The entropy of a Foveon image is higher
    than a Bayer image, all other things being equal.

    So the question is, is the additional information useless...In light
    of how the human eye/brain processes visual information?

    BTW, I know that all other things are not equal in an SD9 v. 10D.
     
    William Wallace, Jan 18, 2004
    #1
    1. Advertisements

  2. William Wallace

    leo Guest

    According to imaging-resource.com's SD10 review, it's 3MP's color, hue and
    resoution are every bit as good as a 6MP Bayer with more details. I wish
    they can up the resolution even more. It has a bit noise than Nikon D100 and
    Canon's 10D & 300D in night shots. Not sure it's the sensor or noise
    reduction algorithm. Nevertheless, the Sigma only lens mount is a big
    deterrent for me.
     
    leo, Jan 18, 2004
    #2
    1. Advertisements

  3. (William Wallace) wrote in
    Nope it does not. A true RGB sensor captures more info than
    a Bayer sensor. The Foveon sensor has, due to its design,
    problems with (1) noise (2) color acuracy.

    Moreover, the SD9/SD10 Foveon sensors lack anti alias filter.
    Thus it creates artefacts that are not true to life.
    A true RGB sensor is better than a Bayer sensor, mainly for its
    better possibilites for post processing of the image. If the image
    shall be used directly from the camera, then the human eye processing
    makes this advantage doubtful.

    The Foveon is not a true RGB sensor - and it has no anti alias filter -
    therefore the comparison is not correct.
    They are very similar. The SD10 is slightly better.


    /Roland
     
    Roland Karlsson, Jan 18, 2004
    #3
  4. William Wallace

    Ray Fischer Guest

    Than what? It captures about, um, 15MB of data while a 6MP camera
    using a Bayer sensor will capute about 9MB of data and a 14MP camera
    with a Bayer sensor will capture about 21MB of data.
    Heh? What do YOU mean by "entropy"?
    Useless? Shrug. Not quite, but a lot of it is redundant. JPEG
    compression works by taking advantage of the fact that images have
    a lot of redundant information.
    Not by a long shot.
     
    Ray Fischer, Jan 19, 2004
    #4
  5. It's up for debate whether or not a Foveon sensor captures more data.
    The RGB sensors have overlapping color ranges and odd hues that need to
    be corrected with heavy processing. Do you really get more color data
    than with Bayer? Maybe. The strengths and weaknesses are different and
    visible with both types of sensors.
    The human eye doesn't perceive color detail well, being that its color
    sensing cells are rather sparse. Any lossy compressor takes advantage
    of that. The Foveon versus Bayer debate is pointless if you're using
    JPEG or an inkjet printer.

    They Foveon sensor is interesting in that it has the potential to
    capture more light than a Bayer sensor. Its downside is that its color
    selectivity can't be tuned to the degree it can with a Bayer sensor.
    Only time will tell whether it, or a derivative technology, will advance
    enough to become the standard type of sensor. Foveon's patent fees will
    also go a long way to determine if their technology is adopted or
    ignored in favor of other technologies.
     
    Kevin McMurtrie, Jan 19, 2004
    #5
  6. William Wallace

    Chris Brown Guest

    In information theory, paradoxically a noisier image has more information.
    That's why they're harder to compress.
     
    Chris Brown, Jan 19, 2004
    #6
  7.  
    George Preddy, Jan 19, 2004
    #7
  8. William Wallace

    Chris Brown Guest

    No it's not. They self evidently capture more data, as the amount of data is
    just a simple count of bits. However, he said "information", which is
    somewhat different.
     
    Chris Brown, Jan 19, 2004
    #8
  9. Right, the biggest mistake is people make is that a;though they realize the
    Foveon design is better sensor for sensor, they fail to realize that the
    Foveon Pro 10M is a 10MP sensor compared to only 6MP, against which it is
    generally priced.

    IOWs, the biggest Foveon advantage isn't the design, it is pure sensor
    count per $ spent.
     
    George Preddy, Jan 19, 2004
    #9
  10. :)

    Now ... let me see ... I have read information theory.
    Does a noisier image contain more information? Hmmmm ...
    not about the subject, but it has more information about
    the noise. I seem to remember that the lecturer we had
    tried to avoid calling calling this "information" and
    that the name "information theory" is slightly misleading.


    /Roland
     
    Roland Karlsson, Jan 19, 2004
    #10
  11. You cannot have it both ways - more data per sensor AND more sensors.
    Sensors per $ is an uninteresting figure.


    /Roland
     
    Roland Karlsson, Jan 19, 2004
    #11
  12.  
    Kevin McMurtrie, Jan 19, 2004
    #12
  13. William Wallace

    Don Stauffer Guest

    I can see the Foveon chip being SOMEWHAT better in resolution. Has
    anyone seen resolution results of the Sigma camera done by the standard
    ISO resolution test? For that matter, anyone seen any tests of ANY
    digicam using the ISO test procedures?
     
    Don Stauffer, Jan 19, 2004
    #13
  14. Yes - I have seen it - a long time ago.

    But - unfortunately the ISO test is not good for digital cameras.

    Take a look here for some serious measurements
    http://www.luminous-landscape.com/tutorials/dq.shtml



    /Roland
     
    Roland Karlsson, Jan 19, 2004
    #14
  15.  
    George Preddy, Jan 19, 2004
    #15
  16. To the ripped off, perhaps.
     
    George Preddy, Jan 19, 2004
    #16
  17. Dpreview includes shots of a standard test chart in every full review
    they do. You can download and compare charts for the SD-9, 10D, and
    1Ds for yourself.

    In the case of the SD-9, you can see that it resolves the pattern with
    the correct number of lines (9) up to 1500 l/ph, then aliases and shows
    patterns of 7 and then 5 lines at 2000 l/ph. The Bayer cameras
    reproduce the pattern correctly up to higher resolution, then give a
    grey blur rather than aliasing.

    Dave
     
    Dave Martindale, Jan 19, 2004
    #18
  18. William Wallace

    CBM Guest

    Question, the chart on the bottom of the page shows the
    Luminance noise graph and the EOS-D60 shows to have lower noise, but to me
    the test shots higher up the page under Sigma SD9 vs. Canon EOS-D60 , shows
    less noise for the SD9 than the D60, what gives.
     
    CBM, Jan 19, 2004
    #19
  19. It demonstrates the relative insignificance of color for detail resolution.
    The Luminance noise is what the Human Visual System perceives as noise, and
    the same applies to resolution. The Green channel has the highest
    contribution for the final impression, and you can see the D60 has a better
    score there.
    That doesn't say color is unimportant all together, but it is a magnitude
    less important.

    Bart
     
    Bart van der Wolf, Jan 19, 2004
    #20
    1. Advertisements

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.