Putting the SD9 "yellow myth" to bed, a follow along RAW demo

Discussion in 'Digital Photography' started by George Preddy, Dec 24, 2003.

  1. Field of view is dependent upon the focal length of the lens and the size of
    the sensor. For a given focal length lens , a larger sensor has a wider
    field of view. That's why the FOV in my old two meg digital with its 8mm
    sensor and 7mm focal length lens is equivalent to a 35 mm lens in a 35 mm
    camera.

    actual resolution of the lens and CCD is measured in line pairs per mm. It
    is independent of the sensor size or target distance. The number of line
    pairs measure along a given dimension is sensor size * resolution in line
    pairs per mm. Resolution is usually CCD limitted. Thoretically it takes 2
    pixels to resolve a line pair but in practice it is closer to 3 pixels per
    line pair. The number of line pairs in a given dimension is approximately
    number of pixels/3. This is further complicated by the fact that many
    reviewers state resolution in lines which is twice the number of line pairs.

    magnification is the ratio of approximately the focal length of the
    lens/distance to object being photgraphed and is dependent upon the focal
    length of lens and the distance of the target. Magnification is also equal
    to the size of the object on the film/actual size of the object. Thus macro
    lenses have a magnification greater than one. The actual size of the object
    on the film is larger than the actual object. Hence you see things that you
    wouldn't see with the naked eye. To acurately measure resolution off of a
    test taget, the reading is only accurate for a particular focal length and
    target distance.

    Everything Dave said makes sense to me.

    I believe the characteristic that George really likes is accutance which is
    a measure of 'sharpness' and not resolution. Acutance is how quickly a line
    transitions from black to white. Many tests have shown that lower
    resolution pictures with high accutance appear sharper to the viewer than
    higher resolution pictures. Lense design is a compromise between designing
    for contrast versus designing for resolution. Most good lenses for
    photography favor contrast. Sharpening works by getting rid of the middle
    values in a transition. Thus if a sharp edge went from black to white over
    5 pixels , the pixel values might be 100 black 75% gray 50% gray 25%gray and
    100% white. After extreme sharpening, those five pixels would be 100%black,
    100% black 50% gray, 100% white, 100% white.

    What makes many people say that the Foveon sharpens it pictures internally
    is that the accutance is higher than a lense would actually produce, there
    for the picture was sharpened.
     
    Gherry Bender, Dec 28, 2003
    1. Advertisements

  2. A full frame sensor doesn't *have* a field of view at all. Only a
    sensor combined with a lens has a field of view.

    An intelligent photographer chooses the lens to give the field of view
    they want for a particular shot. If the sensor size differs between to
    cameras, that means using different lens focal lengths.

    And for testing resolution, resolution test charts are designed to be
    imaged at a particular size, usually with a set of markings that is
    supposed to fill the height of the frame. If you *do* fill the frame
    height as designed, then the markings on the chart can be read directly
    in terms of lines per picture height. If you don't make the chart image
    the correct size, the markings need to be mulitiplied by a scale factor
    to give actual resolution.

    In the case of a digital camera, if you just look at resolution tests at
    100% scale, that only shows you resolution in terms of cycles per pixel
    or pixels per cycle. That's essentially useless when comparing cameras
    with different pixel counts, since what the user cares about is the
    total resolution across the image field of view, not across some group
    of a fixed number of pixels.

    The outbackphoto tests were shot without making the field of view the
    same between the cameras (so the results don't tell you resolution per
    picture height). Nor did they make the targets the same number of
    pixels (so the results cannot be interpreted as resolution per pixel, and
    then multiplied by image size). You can compare how much of the star
    you can see using the star itself for size reference, but the Sigma
    camera was effectively using a longer focal length lens (narrower FOV),
    so it's not surprising it can see more detail. But you can't tell by
    looking how much is due to the unequal test conditions, and how much (if
    any) is due to the supposed superiority of the Sigma's sensor.

    In other words, the test is essentially useless for measuring
    resolution, unless you work out and apply a compensating factor to allow
    for the test conditions. You never mention this.

    Dave
    camera >> and a 50 mm lens, you'll get a certain resolution figure from
    the >> target. If you then put on a 100 mm lens, you'll see detail
    that's >> approximately twice as fine - but over a field that's only
    half as wide >> and high. Has the resolution of the camera increased?
    No - you've just >> magnified the target so the calibrations don't read
    correctly any more. >> The longer lens allows reading finer detail in
    the target, in terms of >> lines per unit distance on the target, but
    the resolution of the >> *camera* has not increased.
    You >> need to change either the lens focal length or the subject
    distance so >> that a specified distance on the target just fills the
    image height or >> width. The outbackphoto test did not do this - it
    used the same focal >> length lens at the same subject distance. Since
    the Sigma's sensor is >> smaller, it saw less of the chart than the
    Canon, and it's not fair to >> compare cropped excerpts of the two
    images under these conditions.
    field >> of view, the targets would have covered more area on the Canon
    sensor, >> and the resolution results would have been better.
    still >clobbered the D60.
     
    Dave Martindale, Dec 28, 2003
    1. Advertisements

  3. George Preddy

    imbsysop Guest

    you'll be looking at the effect of 100% cropped braincells .. LOL !
     
    imbsysop, Dec 28, 2003
  4. Look at those skin tones, gosh she is glowing orange/yellow/green/purple
    neon cheese colors...

    http://www.pbase.com/image/14942333

    No wait, its blue/gold/brown cheese. Oh shoot, I forget, what is the cheese
    color today?
     
    George Preddy, Dec 28, 2003
  5. George Preddy

    JPS Guest

    In message <bsmtbo$41b$>,
    You made me laugh, Steve! The first thing I saw drawn on the screen,
    after the sky, was fake colored trees on top of the hill! As Ralph
    Kramden would say, "you're a regular riot".
    --
     
    JPS, Dec 29, 2003
  6. George Preddy

    Mark M Guest

    Surely you meant to say that to Georgie...
     
    Mark M, Dec 29, 2003
  7. George Preddy

    Mark M Guest

    Sadly, George would apparently not notice odd colors unless they looked as
    he described them above.
     
    Mark M, Dec 29, 2003
  8. George Preddy

    Mark M Guest

    This foolishness is in keeping with his similar inability to see when skin
    looks like cheese, and when real boulders and trees look like plastic.
    George is a man with the unfortunate dillema of trying to adjust images
    without visual discernment.
    Perhaps *to him*, his huge prints look great.
    I have a grandmother whose eyesight is so poor that I needn't buy her pretty
    birthday cards--since she can't appreciate the beauty. Perhaps the Sigma
    folks have found the perfect customer in George.
     
    Mark M, Dec 29, 2003
  9. Ok great, the 1Ds has no FOV.
    Exactly, so no advantage whatsoever to full frame, have it your way. But
    you can't have it both ways. Either there is a point to full frame sensors,
    and therefor you use the exact same lens for resolution tests, or there is
    no point and you change lenses/distance for resolution tests. I'm all for
    it, lets all agree full frame is 100% useless, especially when smaller
    sensors have 25% higher ability to sense full color.
    Right, they actually had the nerve test a Canon with the exact same lens at
    the exact same distance as the Sigma. How rude and unfair! No wonder the
    SD9 blew the 1Ds off the surface of the Earth. It's so obviously cheating.
    Everyone knows there is no advantage at all to having a full frame sensor.

    You just zoom out with a smaller sensor...
    http://www.pbase.com/image/13857420/original
     
    George Preddy, Dec 29, 2003
  10. That's 14MP with the same degree of interpolation as Bayer. PPI is
    meaningless, anyone who even assiociates scaled up PPI with optical res has
    to be an amateur. Maybe you need to shoot something with an optically
    superb camera to understand...

    http://www.pbase.com/image/19169808/original
     
    George Preddy, Dec 29, 2003
  11. The green is so obviously fake. Because everybody knows the SD9 only has
    120% as many green sensors as a 10D...

    http://www.pbase.com/image/17318440/original.jpg

    The greens look like that icky purplish-blue skin tone cheesey things and
    stuff.
     
    George Preddy, Dec 29, 2003
  12. No... it's just you.
     
    George Preddy, Dec 29, 2003
  13. George Preddy

    Mark M Guest

    100% crop simply refers to a section of an image that is presented at full
    resolution...as opposed to a sized-down portion of...or full, but sized-down
    image.
     
    Mark M, Dec 29, 2003
  14. They are that way, can't you see the cheese? Now, these are perfect skin
    tones....

    (nudity warning) http://www.pbase.com/image/21492950

    She's not yellow at all. The 1Ds never colors skin yellow, it just wouldn't
    be right after Canon took $8000 for it. And what fine focus that is, for a
    90% downsize. It's just perfect.
     
    George Preddy, Dec 29, 2003
  15. Now these are fine skin tones...

    http://www.pbase.com/image/21492950

    The blurry old Canon 1Ds doesn't do too bad, why it's almost as sharp as
    Foveon! Oh, wait, that is after it's been downsized to 3MP. I knew the
    11MP $8000 1Ds could be as sharp as a 10MP $600 SD9. It's too bad about
    the blown highlights and pastey white skin tones.
     
    George Preddy, Dec 29, 2003
  16. George Preddy

    imbsysop Guest

    why not just call it a "cut-out" ?? :)
     
    imbsysop, Dec 29, 2003
  17. George Preddy

    Peter Aitken Guest

    Technically perfect perhaps, but otherwise trite, boring, hackneyed, and
    totally without any redeeming aesthetic qualities. Typical of the work of
    those who are primarily concerned with raw images, color depth, and other
    technical matters. There's a pretty clear divide on this group - those who
    enjoy playing with their cameras and computers and those who enjoy creating
    beautiful images.
     
    Peter Aitken, Dec 29, 2003
  18. George Preddy

    Paolo Pizzi Guest

    Pong. :)
     
    Paolo Pizzi, Dec 29, 2003
  19. George Preddy

    Mark M Guest

    That works, but many people still present "cut-outs" that do not offer full
    resolution.
     
    Mark M, Dec 29, 2003
  20. How can it possibly have a field of view when there is no image? There's
    no image until you put a lens on the camera. The lens and sensor size
    together determine the field of view.
    You're not making any sense.

    There are several advantages to full-frame sensors: larger pixels which
    give lower noise, more pixels which gives higher spatial resolution. If
    you're shooting extreme wide-angle shots, you also get more field of
    view, but that's not a consideration for normal and tele shooting; the
    photographer simply chooses whatever lens gives the field of view desired
    at the shooting distance desired.

    When shooting resolution tests, it is normal to adjust for different
    sensor or film format sizes by making the resolution target fill the
    frame. If you want to shoot the test with the same lens to avoid lens
    differences affecting the results, you can either change camera to target
    distance or have multiple versions of the resolution target.

    None of this says that full frame sensors are "useless", but it does say
    that *when performing resolution tests* you should make each camera shoot
    the same area of test target. The outbackphoto tests did not do this,
    so their results cannot be compared in the usual way.
    It's definitely not the right way to do a resolution test if you want the
    results to mean something. Consider the alternative: If you were to
    test a cheap 2 MP digicam with a tiny sensor (example: Canon A200) with
    the same lens at the same distance from the target, the cheap camera
    would have higher resolution than either of the SLRs. It would do this
    because its pixel density is much larger than the SLRs. But the result
    would be meaningless, because the field of view would be 1/10 that of the
    1Ds. The A200 normally uses a 5 mm lens, and with *that* lens it would
    have a field of view similar to a 40 mm lens on the 1Ds - and much lower
    resolution on the test chart. If all cameras are set up for the same
    field of view, this problem is eliminated, and you can compare the real
    ability to capture detail within the same scene.

    Dave
     
    Dave Martindale, Dec 30, 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.