JPG vs TIFF Resolution Test

Discussion in 'Digital Photography' started by Jerry McG, Dec 4, 2003.

  1. Jerry McG

    JPS Guest

    In message <bsc5pe$j83$>,
    No, you pathetic liar. The "6MP" DSLRs made by Canon and Nikon actually
    read light levels at 6+ million 2D locations on the sensor.
    No, you pathetic fool; this is impossible because they collect data in a
    completely different geometrical pattern.

    You still show no understanding of the relative value of luminance
    resolution, and chromatic resolution. If you take two copies of the
    same image with lots of both (such as any digital image downsampled to
    50% or less, and then sharpened), and you separate the chrominance
    channel(s) from the luminance channel, and pixelize (2x2 and/or 3x3)
    chrominance in one, and luminance in the other, the luminance-reduced
    one suffers dramatically, and the chrominance-reduced one looks exactly
    the same, at a 1:1 viewing on the monitor, or in a print. If you take a
    6mp image from a Canon 10D, and do the same with it, you will get
    virtually the same effect. Losing the chrominance resolution has no
    effect, and even a 2x2 pixellation of the luminance channel suffers
    greatly, proving that there really is 6.3MP of useful informatiopn
    there.

    A good Bayer algorithm can start with the premise that color doesn't
    change from pixel-to-pixel (if it really did, and was captured by a
    higher-chrominance-resolution device, you would not see the advantage in
    normal photography). From there, it uses the assumed color, derived by
    interpolation from neighboring pixels, and calculates luminance assuming
    that the two missing colors are in the proportion to produce that color.
    And, of course, it works a lot better than it sounds to you.
    --
     
    JPS, Dec 24, 2003
    1. Advertisements

  2. Jerry McG

    Guest Guest

    but don't forget - 20.56mb 'approaches 100mb.'

    and since foveon 'approaches medium format quality,' by the same
    standard, that image calculates out to be about 11mm film. thats
    smaller than 110 film!!
     
    Guest, Dec 24, 2003
    1. Advertisements

  3. The SD9 produces 14MP images. You should really know that by now.
    According to the pros at Photokina, anyway, after examining the SD9's wall
    sized Durst prints.
     
    George Preddy, Dec 24, 2003
  4. That is correct, but a standard camera like my 300D needs 4 sensors
    to generate full color, as the sensors are arranged in a G-R-G-B
    pattern or something like it. So every sensor only gets some
    of the availble colors.

    So the 300D and friends do interpolate. And normal digiams often
    have dead sensors, which is mapped out and interpolated. That is
    why you can buy cheap cameras with 2nd grade sensors.
     
    Povl H. Pedersen, Dec 25, 2003
  5. Jerry McG

    JPS Guest

    In message <>,
    It is impossible to sample properly with a 2-pixel period. Properly
    sampled images *must* have anti-aliasing, and therefore, the
    hypothetical situation of a single point of light falling on a single
    sensor with the wrong color filter is impossible.
    No one said they didn't. SteveGeorge fabricates stories about other
    people claiming that cameras like the 300D don't interpolate color.

    The 300D samples each color channel in 1.58M locations, and samples
    luminance in 6.3M locations. I haven't seen a single person argue with
    that, except SteveGeorge, who seems to think that 1.58M is the figure
    for all parameters of sampling.
    --
     
    JPS, Dec 26, 2003
  6. Same as all Bayers.
     
    George Preddy, Dec 26, 2003
  7. ....10.3MP of RGB data.
     
    George Preddy, Dec 26, 2003
  8. Uh oh, the word is getting out.
     
    George Preddy, Dec 26, 2003
  9. Different issue only relevant to plastic mosiacs that actually sample color
    using offset locations.

    The issue is...

    The 10D is only a 1.5MP camera, interpolated (yuck) up to 6M ouput pixels.
     
    George Preddy, Dec 26, 2003
  10. Ok, then it's 10.3M samples for Foveon. Take your pick, full color sensors
    or 1/3rd spectrum sensors--you can't win this argument, Foveon has you
    seriously outgunned on every front.
     
    George Preddy, Dec 26, 2003
  11. It works ok. SPP outputs 14MP images by doing precisely the same thing.
     
    George Preddy, Dec 26, 2003
  12. Jerry McG

    JPS Guest

    In message <bsgrjp$8r7$>,
    Nonsense. The issue is very real, and is one of the reasons that only
    English majors working at mags funded by advertising revenue praise the
    SD9, in the allegedly "professional" world. Most people who really know
    the subject are rejecting the camera, because of all its shortcomings.
    That's not in the least bit true. The 10D is a 6MP camera, as the
    dramatic difference between a 6MP image and a 1.58MP image derived from
    it unquestionably proves.


    --
     
    JPS, Dec 27, 2003
  13. Jerry McG

    JPS Guest

    In message <bsgtho$at5$>,
    I've printed SD9 images at the appropriate PPI for 18"x12" prints, and
    10D images, too. The 10D images look like sharp photos, without the
    grain. The SD9 images look like a bunch of blocky pixels stolen from a
    sub-array of a much larger, properly-sampled image. They are sharp, but
    "fuzzy-sharp", with predictable, gridded texture, and are not so much
    "image-sharp".

    Perhaps if you could redesign the camera to take 9 images in a row, and
    shift the sensor by the following amounts:

    0,0 0,0.33 0,0.67 0.33,0 0.33,0.33 0.33,0.67 0.67,0 0.67,0.33, 0.67,0.67

    .... you'd have the data to make a decent-quality 31MP image.
    --
     
    JPS, Dec 27, 2003
  14. Jerry McG

    JPS Guest

    In message <bshdrk$f5$>,
    Only a total moron would see anything "precisely the same" about these.
    The images prove otherwise as well; a 6MP image and a 1.58MP downsampled
    with any method look completely different in a print of the same size.

    Your notion that a camera sees bayer RGB triplets as a single-space
    pixel has no foundation in reality, and shows that all this graphics
    stuff is just a black box to your scrambled brain.
    --
     
    JPS, Dec 27, 2003
  15. Jerry McG

    JPS Guest

    In message <bsgddp$l63$>,
    I didn't say otherwise, what's your point, except to show that you can't
    understand what other people are writing?

    My point was that the data compresses to much; meaning that the data was
    probably not using anywhere near all 12 available bits.

    --
     
    JPS, Dec 27, 2003
  16. Jerry McG

    JPS Guest

    In message <bsgdao$l3j$>,
    Not the same at all. Only a totally braindead person would insist that
    reading light levels at 6.3M locations and interpolating the color is
    the same as reading 1.58M full-color sensors. Color is trivial,
    resolution-wise. Luminance is not trivial at all, and tiny changes in
    luminance are easily seen.
    --
     
    JPS, Dec 27, 2003
  17. You are missing the main degradation of JPEG'ing an image. That happens
    when the very first JPEG'ing takes a 36-bit RAW file carrying several
    million unique colors (each from a 68 billion color palette) and reduces it
    to a few hundred thousand unique colors (each from a 16 million color
    palette).

    Yes, that is lossy.
     
    George Preddy, Dec 27, 2003
  18. You are right about all of that. But a home printer cannot print a 36-bit
    image in 36-bit color, nor can a home monitor display all the color. So
    although there may appear to be no difference at all on a home pc, that
    doesn't mean there isn't a significant difference when developed into a true
    36-bit professional photograph using something like a 36-bit Lightjet or
    Durst Lambda.
     
    George Preddy, Dec 27, 2003
  19. JPEG artifacts are a major problem, but they are far from the biggest
    problem which is conversion from millions of 36-bit colors to thousands of
    24-bit colors, then cutting even that unique color count in half again and
    dithering.
    Obviously the best choice is simply RAW. Then if you want to create a
    professional quality photograph, print a temporary 16-bit TIF to carry all
    the RAW data to a printer that is capable of reproducing 36-bit color (very
    few can). Then delete the huge temporary digital print, and save the RAW
    file with your desired exposure/color/WB settings embedded.
     
    George Preddy, Dec 27, 2003
  20. What images?
     
    George Preddy, Dec 27, 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.