Steve's Digicam's classifies the SD-9 and SD-10 as 10.2MP

Discussion in 'Digital Cameras' started by George Preddy, Oct 27, 2003.

  1. Then you are lying about owning one. What are the the 3 output file size
    choices you see in SPP?

    (see ya)
    George Preddy, Oct 29, 2003
    1. Advertisements

  2. George Preddy

    magnumtrojan Guest


    yet another troll to add to my killfile
    magnumtrojan, Oct 29, 2003
    1. Advertisements

  3. Sorry, couldn't get past the BS. Twice.
    George Preddy, Oct 29, 2003
  4. George Preddy

    MarkH Guest

    Yes, same amount of raw optical data.

    It’s just that you repeatedly lie about how that data is used to create the

    You claim that one green value is thrown away, that is a lie.

    You claim that the data from 4 sensors is combined to make a pixel, that is
    a lie.

    You make up stuff about how Bayer interpolation works which is completely
    untrue, you sir are a very dishonest person.
    MarkH, Oct 29, 2003
  5. I think the main effect of this will be to reduce the credibility of
    those review sites. I hope the site owners realize that.

    Dave Martindale, Nov 1, 2003
  6. And 2X larger spatial separation results in 1/2 the resolution. Are you
    sure you wanted to say that the Foveon has worse resolution? (It does,
    it's true, but admitting that seems so unlike George).

    Dave Martindale, Nov 1, 2003
  7. 10.2/6.3 isn't greater than 1?
    George Preddy, Nov 1, 2003
  8. Wouldn't wanna accurately list the number of sensors on a chip, that could
    be disastrous. Which is not 10.2MP, btw, according to it's:

    Photodetectors: 10.3 million.
    George Preddy, Nov 1, 2003
  9. How many RGB data points are in its 10.3M RGB data point RAW file?
    George Preddy, Nov 1, 2003
    George Preddy, Nov 2, 2003
  11. It is precisely the subject. The answer is:

    3.43 million pixels are required to represent 10.3M individual RGB data
    points in a 10.3MP camera's RAW file. Any more pixels than that represent
    nothing more than wasted bandwidth, pointless up-scaling.
    The SD-9 outputs 13.7MP images. That is, if you start counting from the top
    left corner, and count each and every pixel in the image once, row by row,
    you will reach 13.72 million pixels by the time you hit the bottom right
    corner. That output mode of the SD-9 uses an identical interpolation
    standard to all Bayer cameras (except Fuji, which is a lot worse).

    Thrilled. I've chased image quality through top rated Bayer after top rated
    Bayer, all were dissappointing. Though at the time, much like you, I
    didn't know I had been lied to by Bayer manufactures, and that "6MP"
    actually meant 6 million 1/3rd color pixels before 25% are thrown away for
    lack of complimentary primaries. While Foveon's (only) 3.43MP images meant
    3.43M full color sensors formed using 10.3M 1/3rd color sensors. all with
    complimentary matches resulting in the same optical quality as a 13.7MP
    Bayer, but without the wasted bandwidth of Bayer's 4X upscaling.
    As long as you hope to fool others, I'm going to have to point out that you
    are misdleading them, sorry. Here are the facts, they are not in dispute...
    George Preddy, Nov 2, 2003
  12. George Preddy

    MarkH Guest

    So true
    This is a LIE, one that you have been corrected on many times. Often
    your reply to being corrected on this is very flippant indicating that
    you do know that your statement is a lie but you don’t care. The Foveon
    interpolation is in no way identical to Bayer interpolation. The
    interpolated output by the Sigma is a marketing con and contains no more
    information than the non-interpolated 3.43MPix output. No-one would
    claim such things unless they were lying.
    Also a lie, no values are thrown away and no 4x upscaling. Also Bayer
    manufacturers don’t lie about what their sensors do, the name Bayer
    refers to the RGBG masked grid that is used on those sensors, no one is
    lying about how they work except you.
    It is hard to believe that you could really hope to fool others with YOUR
    lies, they are too transparent.
    Yes, we all know that there are more sensor values from the Foveon
    sensor, all 3 colour values are detected for each pixel. I have yet to
    see anyone on this group dispute this point.

    The evidence is not in that stupid chart, it is in the photos taken. The
    photos clearly show the SD-9 takes sharper pictures than the 6MPix
    competition (though this is likely due to the lack of an anti-aliasing
    filter and more aggressive sharpening in the software) but the colour
    accuracy is sub-par. The most obvious is the yellow skin tones which
    can’t be satisfactorily tweaked in software.

    George, you are one very dishonest person, and the most persistent with
    that dishonesty that I have ever come across.
    MarkH, Nov 2, 2003
  13. All interpolation processes result in no more optical information.
    Smoothing is the only "optical" benefit of both Foveon's (optional) and
    Bayer's (by default) 4X up-scaling.
    How do the extra geen orphans produce accurate colors? And if they do, why
    is there any red and blue sensed at all?

    The real reason there is extra green is very simple, one has no choice with
    a 2x2 scalable grid but to duplicate a color.
    I think you are one of the the first to acknowledge it, most in the group
    once thought a 3MP Bayer (shown in my link too, see the S602's bars) was
    identical in to the Foveon.
    The SD-9 doesn't even set absolute color, it is a RAW shooter. "Popular
    Photography" magazine disagrees with you that its relative color
    reproduction is anything less than stellar, in fact they rate it higher than
    Canon in their latest issue. Same for noise, they rate the SD-9 as less
    noisey than Canon. This is very daring and astute of them, it isn't easy to
    understand that lots of blur and overly aggressive NR is no substitute for
    optical clarity.
    The chart is completely factual, sorry.
    George Preddy, Nov 2, 2003
  14. The question lingers...
    Can you "finish" answering the question? Couthfully?
    George Preddy, Nov 2, 2003
  15. George Preddy

    MarkH Guest

    I can answer that. There are 3.43M RGB data points, or to put it another
    3.43R + 3.43G + 3.43B = 3.43RGB

    My question is this: Why is the colour accuracy so bad is many photos from
    the SD-9 when in theory it should be so much better?
    MarkH, Nov 2, 2003
  16. George Preddy

    MarkH Guest

    Yes interpolation does not produce more optical information, I agree with
    that. Smoothing is the ONLY benefit of Foveon’s interpolation and the
    same interpolation for smoothing can be done on any image.

    But Bayer’s interpolation is not 4x up-scaling at all. It is calculating
    2/3 of the colour values. For each green sensor the red and blue values
    are calculated based on the values of the sensors of the pixel itself and
    its 8 neighbours.
    There are no green orphans! The actual value for each sensor is used for
    the colour value for that pixel corresponding to the colour the sensor is
    detecting. EVERY green sensor value gives the green value of the pixel
    it represents, the red and blue have to be interpolated though.
    I don’t think you understand enough to know the real reason at all.
    Maybe not everyone is acknowledging it, but who is denying it?
    I don’t read Popular Photography and I don’t care what they say. I was
    referring to pictures taken by the SD-9 and put on the web. Specifically
    pictures that you posted links to. The worst one had skin tones about
    right but the shadows on the girls face were yellow, also her ear looked
    VERY yellow alongside where the hairs were. You can try to say that they
    are just bad examples, but YOU were the one that posted the links to
    Yes it is, but that is not what I am disputing here. You are very
    dishonest in your repeated misrepresentation of how Bayer interpolation

    I know some of the people that have argues with you have also been a
    little inaccurate, but they have mostly understood and described the
    process as accurately as they knew it. Some say that the luminance value
    of each pixel is known, obviously only the luminance value of 1 colour is
    KNOWN at each point (pixel) the other 2 colours have to be calculated.
    If it wasn’t for how well the values were calculated in all but rigged
    tests deliberately designed to trip up the Bayers, I wouldn’t be happy
    with my Bayer results. Luckily there is enough information known to
    interpolate the rest with impressive accuracy.

    I usually shoot JPG with my 10D with the sharpness set to +3, how the
    pictures come out of the camera is how I use them, I really can’t be
    bothered wasting hours in Photoshop CS. I am very happy with the
    sharpness and colour accuracy of my pictures even viewing them at 100% or
    more on my 19" monitor.
    MarkH, Nov 3, 2003
  17. SNIP
    Because it is hard to separate the RGB spectral absorptions caused by
    silicon in distinct bands with a large dynamic range. Separation requires
    subtraction, and that decreases the already limited Signal/Noise ratio.

    Bart van der Wolf, Nov 3, 2003
  18. You win! Congratulations on being the first to give the correct answer.
    WB. Namely, AWB

    Seriously, that's it. I've played around a lot with this, to the tune of
    4500 SD-9 photos, WB problems are primarily what people report as poor color
    rendition. Look at's SD-9 complaint of too-blue skies, the
    gray gondolas are clearly too blue as well, ref them to neutral gray and
    presto the image is beautiful, including the deep sky color (Bayers have a
    heck of a time with acurately reproducing blue skies--hence my large
    collection of 1Bs and polarizers that are no longer needed to capture
    naturally beatiful blue skies with digital). Anyway, AWB is definitely
    skewed toward blue neutral, which produces yellow skin tones. Solution is
    simple, when using AWB correct the WB manually or automatically using the
    neutral dropper, the SD-9 never sets WB permanently. Or alternatively,
    shoot a more appropriate preset WB, or better still shoot a custom WB.
    Given SPP's ability to apply a WB setting to a previous shot, it makes no
    sense at all not to shoot a neutral card after you capture a great shot.
    Why not get the WB perfect?

    Another issue is beta firmware, Askeys review was done entirely with beta
    firmware and never updated, even though he notes that most of what he
    reported as problems were not there in rev 1.1, which was available at
    public release. Also of note, this time Askey got dissed by Sigma with
    their SD-10, he had to link another site for "his" pre-release review. Good
    on Sigma, that is exactly what they should do to pushy Canon sites.
    George Preddy, Nov 3, 2003
  19. Recycling already used neighbors adds no optical value. It might guess the
    value a little better, but the optical limit is the same and is set by the
    number of complete RGB sets.
    But you can't seem to explain the magic involved with having 2 green pixels
    for every 1 red and 1 blue, then arriving at appropriately weighted green
    channel values in teh final pixel, while not throwing away any green
    Plenty. I don't think anyone here realized that 3.4MP is all it takes to
    represent 10.3M (properly located) RGB sensor values. Bayer's up-scaling is
    nothing more than embedding loads of unecessary junk in the final
    interpolated output file, caused by the need to combine dislocated RG(G)B
    sets using min processor power.
    Maybe you should start?
    How do you suppose only Foveon ears turn yellow? How silly. Could it be
    a sunset shot? And to be honest, I'm working on a laptop for the time
    being, the display is poor to midland at best (high end model), and since
    the SD-9 never sets color, my settings could definitely be off depending on
    one's monitor choice. Bottom line, if you see an SD-9 image with what you
    think is improper color there is only one person to blame, since the SD-9
    doesn't set color. It is a RAW shooter.

    Better, why don't you shoot both for a while instead of interpolating your
    truth from others' guesses? Or if you can't get a trial, do the next best
    thing and download the software and a few sample RAW files. It'll change
    your mind quickly about what others guess.
    If it worked... :)
    I like the 10D a lot for overall value, but with only 6MP-interpolated, and
    very soft (nice word for blurry) at that, it has an optical color resolution
    weakness I personally couldn't overlook (B&W target tests are meaningless
    when comparing Bayer to Foveon, which is why they are used).
    6MP-interpolated isn't enough for jump-off-the-page sharp, professional
    quality images IMO. USM is a poor substitute for optics.

    These lighbulbs are optics...

    Would you mind linking a few non-sharpened samples, full size?
    George Preddy, Nov 3, 2003
  20. George Preddy

    MarkH Guest

    It is not done for optical value, the goal is to make the best possible
    guess for the 2 colours not sampled at that pixel.

    Bayer does not use complete RGB sets.
    There is no magic and I have explained exactly how it works. The green
    value is guessed for half the pixels, the red for 3/4 and the blue for 3/4.
    More measured green values means there are less guessed green values for
    the same number of total green values as total red and total blue (6
    million of each).

    I think that most serious D-SLR user DID realise that, especially the ones
    that read the reviews.
    Why, my hobby is taking photos, not reading magazines. There are plenty of
    websites that have more in-depth and accurate information then you find in
    those magazines.
    I have no interest is switching to another lens system, my lenses wont work
    with it. I am happy with my current camera. The images that I am
    referring to can’t be used to argue with anyone except you, but since you
    were the one using them as examples of how good the SD-9 is then they
    cannot be dismissed by you.
    The smiley is supposed to indicate a joke, so obviously you are agreeing
    the interpolation for Bayer works fine. Thanks for that concession.

    Err, there are some really messed up colour inaccuracies in that shot!
    Nice picture overall and you have to look closely to see the colour
    fringing. But look at the full 3.43MP image at the peoples legs on the
    left side, see the 2 pairs of legs that are really yellow? The Sky looks
    blue the clouds look white, it seems that the white balance is correct, but
    some of the skin tones are rendered in what I’ll call Foveon yellow.
    Yes I would mind. The JPGs straight out of the camera are 2 to 4MB is size
    and I only use them for printing. For the web I downsize 9 pixels to 1 to
    produce a large but manageable image. Not every country has cheap high
    speed internet like the states. My ISP lets me have 10MB of webspace for
    free and I only connect at 128Kbs.

    But here’s a crop that has had nothing done to it:
    Cropped in Photoshop but nothing else, no resize, no sharpen, no
    autolevels. This is exactly how it came out of the camera.
    MarkH, Nov 3, 2003
    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.