flawed megapixel experiment

Discussion in 'Digital Photography' started by Bucky, Nov 22, 2006.

  1. Bucky

    Bucky Guest

    NY Times tech columnist David Pogue tried to prove that 13mp does not
    yield better results than 5mp, but in a flawed experiment. What was his
    technique? He took a 13mp photo, then downsized it to 8mp and 5mp. He
    printed all three on a 16x24 inch poster and had people try to figure
    out which was which--they couldn't.

    Pogue insists that his test is valid because he wanted to isolate
    megapixel as the sole factor (rather than optics and electronics).
    Anyways, read his article and see if you can convince him of his flawed
    experiment.

    http://pogue.blogs.nytimes.com/2006/11/21/21pogues-posts-2/
     
    Bucky, Nov 22, 2006
    #1
    1. Advertisements

  2. Bucky

    Rudy Benner Guest

    Who says the experiment was flawed?
     
    Rudy Benner, Nov 22, 2006
    #2
    1. Advertisements

  3. Bucky

    Bucky Guest

    Plenty of the comments on his blogs.

    His article implied that a 5mp camera will produce an equivalent 16x24
    inch print as a 13mp camera, so don't bother getting a 13mp camera.

    When actually all that his experiment implied was that a 5mp image
    downsized from a 13mp camera will produce an equivalent 16x24 inch
    print as a 13mp image from the same camera. There's a big difference
    between the 2 conclusions!
     
    Bucky, Nov 22, 2006
    #3
  4. There are lots of factors to consider, but fundamentally I find that when
    testing high pixel density sensor cameras, resolution is impressive at the
    lowest ISO setting, but to keep noise under control at higher ISO settings
    you will lose resolution. High pixel density sensors are less sensitive and
    this reduces their dynamic range, so unless shooting conditions are ideal,
    you will get poorer shadows and highlights, details that will show up in
    large prints.

    In comparisons I have done, comparing, say, ISO 800 with a 5-6MP DSLR with a
    ISO 800 shot from 10MP DSLR (both similar APS class sensor sizes) will show
    pretty similar resolving capability, but the 10MP cameras will be much
    stronger at 100 or 200 ISO.

    Ian

    Digital Photography Now
    http://dpnow.com

    *** Extra 40MB of storage space on DPNow's free photo gallery until the end
    of November, don't miss it while it's there! http://galleries.dpnow.com
     
    Digital Photography Now, Nov 22, 2006
    #4
  5. Bucky

    bugbear Guest

    It certainly justifies a narrower conclusion
    than the one he draws - the NUMBER of pixels
    needed for a print is lower than many people
    think.

    However, the number of pixel sensors in a camera
    is part of a complex set of factors in the
    information gathering capability of that camera.

    So his larger conclusion is NOT justified.

    In the current market, it strikes me
    that P&S cameras *should* have a sensor
    with around 5-6 Mpixels on a moderately
    large sensor.

    This would give "enough" resolution
    (as per the article) but with improved
    ISO sensitivity and/or noise.

    BugBear
     
    bugbear, Nov 22, 2006
    #5
  6. His article implied that a 5mp camera will produce an equivalent 16x24It's this piece of logic where I see the flaw. Even if the prints obtained
    by a 5mp camera are just as good as those produced by a 13mp, the latter
    gives you more flexibility to crop and zoom in post processing.

    Whether or not this is an advantage depends upon how you use your camera.
    Many people won't ever do any post processing but, equally, many will.

    So for some people it may indeed be true that there's no point getting 13mp
    but it's an over generalisation to advise all users that it's not worth
    while.

    Having said that, I am firmly in the camp who believe megapixels are
    overrated. Whilst they might be on my list of purchase criteria they
    certainly wouldn't be near the top. For a long time I used a 2.1Mp camera.
    It had a 10x stabilised optical zoom so I was nearly always able to frame my
    shots exactly as I wanted and hardly ever did any retrospective cropping.
    Never once did I yearn for extra megapixels.

    The trouble with megapixels is that you can count them, so the marketing men
    have latched onto this figure as a universal index of how good a camera is.
    Unfortunately a large section of the buying public have swallowed this
    concept hook line and sinker.

    Recently I upgraded to a new camera because the old one was too bulky. My
    latest purchase boasts 6mp - not a lot by moderns standards but more than
    enough for my modest needs. I have recently discovered an unforseen
    disadvantage of all those extra megapixels. With my old camera, all the
    photos I wanted to keep in any one year fitted, conveniently, onto a single
    CD. I therefore have a cupboard of CDs with labels such as "photos 2005"
    etc. With my new camera they don't. It's too much hassle to have to split
    the year up so suddenly I am in the market for a DVD recorder. Yet more
    expense!

    Keith
     
    Keith Sheppard, Nov 22, 2006
    #6
  7. I read that post, and a bunch of comments, yesterday. Did he ever get
    around to saying what camera was used? If it was an DSLR, what lens?
    You have to wonder whether his test photo was resolution-limited by the
    optics rather than the sensor.

    -dms
     
    Daniel Silevitch, Nov 22, 2006
    #7
  8. Bucky

    Scott W Guest

    I have seen people do this before. Of course the flaw is that a 13mp
    image down sized to 8 will be much sharper then an 8mp photo taken with
    an 8 MP camera. Almost any digital camera image and be down sample to
    about 70% to 75% of it starting pixels count and loose almost no
    detail.

    Problem number 2 is portraits tend to need far fewer pixels then
    landscape photos, he should have done a scene with a lot more detail in
    it.

    Problem number 3 is we have no way to know how well focused the photo
    was, did he use a f/1.4 lens wide open. The difference between 13 and
    5 will not be huge when the 5 is a down sampling of the 13 but it
    should be clearly visible, but only if the 13MP was sharp to begin
    with.


    Scott
     
    Scott W, Nov 22, 2006
    #8
  9. I doubt it. I recently resolution tested a 10MP Casio Exilim EX-Z1000
    compact p&s and it out-resolved the ISO resolution test target at ISO 100.
    At higher ISO the softening of the image to suppress noise did severely
    compromise the resolving power though.

    Ian

    Digital Photography Now
    http://dpnow.com

    *** Extra 40MB of storage space on DPNow's free photo gallery until the end
    of November, don't miss it while it's there! http://galleries.dpnow.com
     
    Digital Photography Now, Nov 22, 2006
    #9
  10. Bucky

    bugbear Guest

    On a tangentially related point, has
    anyone else noticed that recent TV
    programs recorded in HD look better -
    even on normal TV's ?

    BugBear
     
    bugbear, Nov 22, 2006
    #10
  11. Keith Sheppard wrote:
    snip
    I consider pixels near the top of my list, but would prefer that
    instead of mpix they specify actual measured resolution. For a long
    time I argued about that, but I figure feint hope of that ever coming
    to pass. At least the number of pixels sets the MAX resolution.

    Sort of reminds me of early hi-fi advertising, when they would give a
    frequency response range, neglecting to say that response was down 20db
    at those points :)

    And there certainly are other parameters of equal importance. One that
    doesn't seem to get much attention is flare performance.
     
    Don Stauffer in Minnesota, Nov 22, 2006
    #11
  12. Hey, that's a good idea, but I would modify it to measure resolution at
    different ISO levels. It would also have to be independent of lens
    resolution somehow, unless the lens was fixed, like a Sony DSC-R1, for
    example.

    Ian

    Digital Photography Now
    http://dpnow.com

    *** Extra 40MB of storage space on DPNow's free photo gallery until the end
    of November, don't miss it while it's there! http://galleries.dpnow.com
     
    Digital Photography Now, Nov 22, 2006
    #12
  13. Bucky

    Bill Funk Guest

    How many cameras are there on the market that have (close to) 13MP?
    Only 2 that I know of; the Nikon D2x and the Canon 5d.
    Any others?
    According to Dpreview, there are no currently marketed non-DSLRs with
    13MP.
    So, it would seem he used one fo those two cameras. What lens he used
    is unknown.
    The real question, though, is this: what currently marketed 5MP camera
    does he think will match the output of either of those DSLRs?
    I proposed a test: he brings his 13MP camera to the Eifel Tower in Las
    Vegas; I'll being my two 5MP cameras (Canon S2IS and Lumix FX01), and
    we'll shoot the nightime Strip. He can have the resulting images
    printed, and hang them up, and ask people to compare.
    IMO, the results will show a marked difference betwen the 13MP camera
    and the 5MP cameras.
     
    Bill Funk, Nov 22, 2006
    #13
  14. Bucky

    Bucky Guest

    No, he didn't, but it almost certainly was a DSLR. Not too many 13mp
    cameras out there.
     
    Bucky, Nov 22, 2006
    #14
  15. Bucky

    Bucky Guest

    Well, he's trying to isolate MP. But to truly do that, you would need 2
    identical cameras except one has a 13mp CCD, the other 5mp CCD.
     
    Bucky, Nov 22, 2006
    #15
  16. If he wants to isolate the effect of having more megapixels, he could
    start by shooting a 13 MP image with one camera, using a tripod, reduce
    that image to 5 MP, and then print both 5 MP and 13 MP image at 16x24.
    I bet people would be able to see a difference.

    And an actual 5 MP camera could only be worse than his 5 MP downsampled
    image, not better, so this experiment would set a sort of lower bound on
    the possible difference.

    What he actually demonstrated is that if a 5 MP image will make you
    happy, a 13 MP camera is overkill. Wow - nobody would be able to figure
    that out for themselves, would they?

    Dave
     
    Dave Martindale, Nov 22, 2006
    #16
  17. Are you sure they're "recorded in HD"? Most high-quality TV (e.g.
    dramas) are shot on film, then transferred to video. A lot are
    transferred to HD video for future value, even if they aren't broadcast
    in HD now.

    What you're probably seeing is that when the transfer is done to HD, and
    all the editing and effects are done in HD, you get a cleaner result
    (even after final downsampling to standard definition) than if you did
    all the editing in SD.

    Dave
     
    Dave Martindale, Nov 22, 2006
    #17
  18. Bucky

    Bill Funk Guest

    The point is, you just can't isolate the megapixels and come up with
    anything that's relevant to consumers.
    We must buy *CAMERAS*, not sensors. There are no identical cameras
    such that one has 5MP and the other has 13MP.
    We must deal with reality; the reality is that Pogue's test showed
    *only* that his 13MP image printed equally well when printed at 13MP,
    8MP, and 5MP. He got to choose the image, he got to choose the
    printing method, and he got to choose the original premise.
    However, this doesn't deal with the realities that anyone who's
    looking to buy a camera faces. There are no 13MP P&S cameras. There
    are no (currently marketed) 5MP DSLRs.
    If he had wanted to compare 13MP DSLRs with 8MP DSLRs, it would have
    been a far more valid comparison, but only if he'd actually used those
    cameras, instead of only the 13MP camera. And, of course, since there
    are more than one 8MP DSLR, his conclusions would have only been valid
    for the one he'd used, assuming that he'd used the same lens on each
    camera.
    But, since the market doesn't provide the cameras he'd need to prove
    his premise, we must instead come to the conclusion that he can't
    isolate only the megapixel count to prove his premise; it isn't
    possible. The consumer simply can't make a decision on which camera to
    buy based only on megapixel count. (Well, obviously, one *can*, but
    such a decision would be foolish.)
    The end result is that his test doesn't reflect the reality that he
    sets out to prove.
     
    Bill Funk, Nov 22, 2006
    #18
  19. Bucky

    Hebee Jeebes Guest

    I don't think it is flawed. I think it is one of many ways to try to prove
    this. In the end since it all comes down to personal tastes none of it
    proves very much.

    I for one has stopped updating to higher resolution camera because I want
    more detail. I do it now because I want more cropping capability and still
    be able to make a decent sized print at a decent resolution.

    R
     
    Hebee Jeebes, Nov 23, 2006
    #19
  20. Bucky

    Ray Fischer Guest

    One could argue that it would be less sharp because of the effects of
    the downsampling.
    That's nonsense.
    There's plenty of fabric with lots of detail.
    Which might be part of what he's trying to show.

    Keep in mind that the difference between 5mp and 13mp really is not
    that much. In a linear row of pixels it doesn't even double the
    detail. You're only replacing one pixel in the photo with less than
    three, and that really doesn't give a lot more detail.
     
    Ray Fischer, Nov 23, 2006
    #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.