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. That's the whole point, make a print at the same size (ppi) as a 10D one and
    look for the differences.

    What does look strange, is the out-of-focus hairs.
    A single hair changes in 2 or 3 parallel hairs, looks odd.

    Bart
     
    Bart van der Wolf, Nov 6, 2003
    #61
    1. Advertisements

  2. SNIP
    400%.

    Which is great. It's even better than human vision which has much poorer
    color resolution.

    Bart
     
    Bart van der Wolf, Nov 6, 2003
    #62
    1. Advertisements

  3. The heavy sharpening artifacts in the 10D picture truly disqualify it as
    anything but a medium-to-low prosumer quality image. Even the residual
    sharpness is probably lower than is available at the prosumer level as well.
    Once again, you are confusing one's monitor/color prefence with the camera.
    The SD-9 is a RAW shooter, any skin tone (from 68 billion choices,
    literally) is settable. Also, it isn't possible to adjust WB other than
    from a RAW image, JPEG fixes the gamut (to 1/4096th actual size, and skewed)
    so once a JPEG is produced WB is permanently set or permanently ruined,
    depending on one's preference. You can hack on it with Photoshop, but
    overall color gets unrecoverably twisted. That's why JPEG is suitable for
    prosumer work only, or possibly forced low bandwidth pro work when medium to
    low caliber image quality is sufficient.

    Here are some more versions of that picture for your perusal, all are
    originals, see if any of the WBs better suit your monitor/taste. If you can
    home in on one, I'll produce the same number again, centered around that one
    for you with more subtle variations. And if you can pick one of those we'll
    start to tinker with saturation to see if we can make it fit your taste...

    http://www.pbase.com/image/23043466

    Also note, the RAW photo was severely overexposed by accident (super quick
    snap [I suppose thats obvious from the bubble], the SD-9's AF performance is
    incredible, almost no ambient light too) which doesn't help the overall
    color. Bad photographer, incredible camera.
    The SD-9 is the well out in front as the best DSLR for portraits, IMO. The
    clarity, sharpness, and detail it captures combined with its silky film like
    images really can't be matched using a Bayer's fairly erratic color
    interpolation and artifacting. The only drawback is that seeing every skin
    cell can be somewhat unflattering in certain cases, but its easy to hide
    detail in post processing, impossible to create it.
     
    George Preddy, Nov 6, 2003
    #63
  4. Not sure what you mean, but have you ever tried to photograph a purring cat?
    It ain't easy. 6MP-interpolated is nice in that way because you never need
    to worry about sharpness subtleties, can't see 'em. Saves big bucks on
    lenses too, assuming one is realistic about what the sensor is really
    capable of.

    I think examining areas surrounding the whiskers/lashes on these cat images
    says a lot. The 10D is so aritifact prone, there is no way it that image
    could ever be confused with film, or pro caliber, it has low (very low,
    honestly) quality digital written all over it. The Foveon has at least
    double, probably 3X the optical resolution, and has no artifacts to speak of
    (other than the JPEG artifacts for internet media) which could immediately
    distinguish it from film.

    Also of note, this is the lowest quality image (24-bit JPEG) an SD-9 can
    produce. Keeping the original 3 million plus + unique colors while
    preserving the original 68 billion color colorspace, rendered on high end
    silver halide paper using a true non-color-interpolated (Bayer cameras need
    not apply) photographic emulsion development process, gives an enormous
    boost to image quality over a 24-bit monitor rendition (a 32-bit video
    setting is still 24-bit color) of a lossy artifact prone JPEG reduced to
    about 300,000 unique colors, total junk, and those choosen from 1/4096th of
    the original RAW colorspace. The difference between the actual image, and
    what you see on screen, is night and day.
     
    George Preddy, Nov 6, 2003
    #64
  5. Now if we only lived in a 2 color world, the 10D would at least do ok. The
    SD-9 still outperforms it in B&W, and the 10D costs about 50% more, but it
    would be closer.
     
    George Preddy, Nov 6, 2003
    #65
  6. George Preddy

    MarkH Guest

    You just don’t understand, I am NOT referring to white balance and it is
    not my monitor. I am reading the colour values with the eyedropper tool
    in Photoshop, they are not influenced by monitor settings. White balance
    will not affect yellow artefacts on the ear, white balance affects the
    entire image. I am not saying that the colour is wrong in the entire
    image!
    Eeeeek, she looks positively ashen, take her to a hospital. The colour
    looked better in the other version (except for the yellow shadow and
    yellow fringing on her ear).
    Maybe it is just in some bad lighting situations where these problems
    occur?

    Unfortunately the biggest problem is you. You can’t tell the difference
    between white balance and colour anomalies. Tweaking white balance and
    colour saturation will not fix the problem. The only fix is to retake
    the picture in better light and see if it comes out better.

    No camera will take perfect photos under all lighting conditions, only a
    crazed zealot would refuse to believe that a picture may not have worked
    properly.

    From my camera the majority of failures are due to camera shake in low
    light causing blurred pictures. Unfortunately the obvious cure is a
    Canon 70-200 f2.8 IS lens, It’ll probably be at least a year before I can
    afford something like that.

    For the SD-9 the failures I have seen are probably caused by tricky
    lighting conditions, sometimes you need to reshoot, get over it.
     
    MarkH, Nov 6, 2003
    #66
  7. We didn't get to color yet. That's adjustable too. Have a preference?
    It will indeed. Most think like you, and slowly come to realize it is
    indeed a simple matter of color preference, totally tunable. of course no
    camera is perfect, but the SD-9 is no worse than the other DSLRs, and
    actually quite excellent on skin.
    I just showed you its RAW development preference, but you appear to have
    opted out of the demonstration perhaps to save face.
    Even better yet, shoot a neutral card right after a great shot, then
    generate, save, and back fill the custom WB setting. The SD-9's custom WB
    is superb, and its not even necessary to set in advance.
     
    George Preddy, Nov 6, 2003
    #67
  8. Forget the numbers, it's the picture that counts. The borderline cases
    are not that interesting anyway, it looks messy on either camera.
    Look at the vertical resolution bar, around the 14, or 13 mark if you
    want. This is a resolution that is well within both cameras
    capabilities according to the numbers. Answer this: on which camera is
    it easiest to count the number of lines?

    Or look at the horizontal bar around the 12 mark. Which camera best
    shows nice well defined lines a regular distance apart? You claim the
    Sigma has three times the resolution of 10d/D100, is that consistent
    with what you see on _these_ pictures?

    You seemed to have missed this, so I'll repeat it: There is some
    banding perpendicular to the lines in some of patterns on the SD-9
    chart, is that because of the superior detail of the SD-9?

    Oh, and on the tall vertical bar with horisontal lines and a scale
    from 1 to 10 the lines in the regions below 8 seems to be of uneven
    thickness? But on the D100 image, they all seem to be of the same
    thickness (And same distance apart.) Is that because the D100 doesn't
    pick up the unevenness in the printed chart due to it's non-sharpness?
    How is it physically possible for a camera to overachieve their
    optical resolution (3 to 4 times no less)? Is the firmware psychic?
     
    Asbjørn Bjørnstad, Nov 7, 2003
    #68
  9. Bayer sensors are all 1/3rd color. They can all indicate black
    individually, but it takes 3 (4 really as 1 green is wasted) to resolve a
    color. See http://www.outbackphoto.com/artofraw/raw_05/essay.html

    The SD-9 has roughly 2.3X the optical resolution of the Canon 10D in full
    color.
     
    George Preddy, Nov 7, 2003
    #69
  10. George Preddy

    MarkH Guest

    /me shakes head.

    George you just don’t understand what I am saying, this is not a white
    balance issue. I have said all I can and you don’t get it.

    I don’t think I’ll bother anymore.


    Yeah, yeah, I know I shouldn’t have starting replying in the first place.
    Never argue with idiots, they drag you down to their level, then beat you
    with experience.
     
    MarkH, Nov 7, 2003
    #70
  11. You didn't answer the question.
    I know how a bayer sensor works. Have you thought about what would
    happen if the black/white edge goes halfway through your imaginary 2x2
    sensor group? (say the red and one green sensor sees the white part.)
    What color would you see? Can you point to such an effect along a bw
    edge in the bw chart?

    I notice you didn't answer a single question I asked about the image
    quality in the link you posted. Can't make yourself admit the Sigma
    creates artifacts (Banding and uneven thickness and distance between
    close lines.) or is it too hard to say the quality of the D100 is
    better (Easier to count the number of lines around the 12 mark of the
    resolution bars)?
    And here I thought both cameras took color images all the time.
     
    Asbjørn Bjørnstad, Nov 7, 2003
    #71
  12. It is absolutely a settable WB and color issue. SD-9 skin tones are superb,
    you just need something more that a "full-auto" approach to photography,
    which is mostly what you see on the web. I demonstrated that you (only
    half way done) but you opted out when you saw the writing on the wall.
     
    George Preddy, Nov 7, 2003
    #72
  13. B&W targets are meaningless. Obviously no camera is perfect, but the SD-9
    has 2.3X the 10D's color resolution. Is that perfect? No.
    They do, which is why it matters and B&W targets don't.
     
    George Preddy, Nov 7, 2003
    #73
  14. George Preddy

    MarkH Guest

    Ah, George. I can see you are able to win any argument. Of course it is
    by using the technique of putting your fingers in your ears and loudly
    yelling lalalalalala lalalala my camera is better lalalala lalalalalalala,
    when the other person walks away shaking their head you pull your fingers
    out of your ears and say "they knew I was right and they opted out of the
    argument, I win!". You then have the satisfied smile of the mentally
    deranged.

    Whatever George, you win.
     
    MarkH, Nov 7, 2003
    #74
  15. So are you going to see if the SD-9 skin tones suit your preferences
    perfectly or not? They will, but at this point you already know that, don't
    you?
     
    George Preddy, Nov 7, 2003
    #75
  16. George Preddy

    Tony Spadaro Guest

    Probably the only fun an SD-9 user can have since he's stuck with such a
    pathetic camera.


     
    Tony Spadaro, Nov 8, 2003
    #76
  17. ....

    George:

    I realize that you have a lot to keep up trying to respond to every message, but
    I've tried to explain this to you at least twice before in two other threads. Did you
    not bother to read the explanation? Or are you being disingenious?

    ....
    ....

    Bayer sensors don't work the way you are implying they do. You seem to believe
    that each block of four sensors (RGGB) is used to resolve a pixel. (Hence your
    wasted green comment.) Try to get this through your head: Every sensor in a Bayer
    array is the basis for 1 pixel. Every R, G, and B produces a pixel. By design, twice
    as many pixels are based on the G sensors, as either of the other two sensors.
    Clearly, each sensor provides only a part of the information for a its pixel. The
    remaining
    information is calculated from the surrounding sensors (at least 8).

    You cannot equate an RGGB block (which has no importance than to show how the
    sensors are laid out) with the "triple sensor" in a Foveon array. In the Foveon array,
    each pixel is based on one "triple sensor". So, the real comparison is between an
    array of n "single sensors" in a Bayer chip, with m "triple sensors" in a Foveon chip.
    The Bayer chip produces n pixels based on (relatively) accurate luminance information,
    and partial chromatic information. The Foveon chip produces m pixels based on
    relatively accurate luminance and chromatic information.

    Assuming two relatively well designed cameras: Clearly if m >= n, the Foveon camera
    should produce the better picture. However, if m < n, at some point the greater
    quantity
    of less accurate samples from the Bayer array will produce a better image than the
    fewer, but more accurate samples from the Foveon array. Visual examination of samples
    from current cameras seems to suggest that the point is somewhere in the 4-5MP range.


    --
    Dan (Woj...) dmaster (at) lucent (dot) com

    "I believe in the kingdom come / Then all the colours will bleed into one
    Bleed into one / But yes, I'm still running
    You broke the bonds and you loosed the chains
    Carried the cross of my shame / Of my shame / You know I believe it"
     
    Dan Wojciechowski, Nov 10, 2003
    #77
  18. It not a group its an optical limit. A camera is ultimately limited by the
    amount of information it sees. If, as in the case of the 10D it only sees
    1.58MP worth of color information, it can interpolate and compine the same
    sensors over and over again in all srts of different output pixels, but it
    will never create more optical data that it sees.
    Exactly why only optical color resolution is important when discussing color
    digital cameras. A Bayer sensors does the best if the image is pure black
    (no color) as each individual pixel is, in that one special case, 100%
    accurate, instead of being nothing more than a guess.
     
    George Preddy, Nov 11, 2003
    #78
  19. George Preddy

    Guest Guest

    The Sigma cameras only sample 3.4 million spatial sampes. The competitors
    sample 6 million spatial samples.
     
    Guest, Nov 11, 2003
    #79
  20. No they don't, they take 6M 1/3rd samples.
     
    George Preddy, Nov 11, 2003
    #80
    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.