Re: 24 megapixels versus 12, substantial diff.

Discussion in 'Digital Photography' started by Scott W, Feb 17, 2009.

  1. Scott W

    Scott W Guest

    On Feb 17, 1:11 pm, Alan Browne <>
    wrote:
    > Rich wrote:
    > > Simulation of D3x against a D300's output.  I shot 4 frames with the D300,
    > > combined and cropped them down to 24 (actually 26.5) megapixels and also
    > > took one shot for 12 megapixels.  The detail increase is more noticeable
    > > than I'd have expected.  I figure this is pretty much like going to the
    > > D3x, at least at lower ISOs.  Images look best at top screen resolution (at
    > > least 1680 x 1050).  Likely a true 24 meg image from the D3x would yield
    > > more detail still, owing to the fact you aren't dealing with lens edge
    > > quality in the middle of the frame from combining four shots.

    >
    > > 12 Megapixels:
    > >http://www.pbase.com/andersonrm/image/109265567

    >
    > > 24 Megapixels:
    > >http://www.pbase.com/andersonrm/image/109265645

    >
    > 1/res_out = 1/res_lens + 1/lens_sensor.
    >
    > eg: you're neglecting lens resolution losses.


    I think you meant to write 1/res_out = 1/res_len + 1/res_sensor, you
    had lens_sensor.

    But the real equation is closer to 1/res_out = sqrt(1/res_len^2+1/
    res_sensor^2)

    Still you are correct that the lens needs to be taken in to account,
    and can be the limiting factor.

    Scott
    Scott W, Feb 17, 2009
    #1
    1. Advertising

  2. Alfred Molon wrote:
    []
    > Ehmmm... a lens does not have a resolution in megapixel. Both formulas
    > make no sense.


    Alfred,

    It makes sense if you use the same spatial units, whether that be microns
    or megapixels, and then use the appropriate formula. But you are right
    in that using one figure for lens performance may simplify things too
    much.

    David
    David J Taylor, Feb 18, 2009
    #2
    1. Advertising

  3. Alan Browne wrote:
    > David J Taylor wrote:
    >> Alfred Molon wrote:
    >> []
    >>> Ehmmm... a lens does not have a resolution in megapixel. Both
    >>> formulas make no sense.

    >>
    >> Alfred,
    >>
    >> It makes sense if you use the same spatial units, whether that be
    >> microns or megapixels.

    >
    > Megapixels is a count over a surface area, so not too useful - lp/mm
    > would be the 'typical' unit.


    In both cases one is simplifying things by assuming a single linear
    measure (pixel effective size or lens point-spread function), and then
    applying an RMS-style addition of these values to produce an effective
    system resolution. This may work well if all the system MTFs are
    Gaussian, but not so well otherwise. For a given sensor size, you could
    express the lens resolution as "megapixels", or you could express the
    sensor resolution as line-pairs per mm. All approximations, of course.

    David
    David J Taylor, Feb 18, 2009
    #3
  4. Alfred Molon wrote:
    > In article <RsRml.36722$>, David J
    > Taylor says...
    >
    >> It makes sense if you use the same spatial units, whether that be
    >> microns or megapixels, and then use the appropriate formula. But
    >> you are right in that using one figure for lens performance may
    >> simplify things too much.

    >
    > The problem is that a lens does not have a resolution - it has a
    > certain MTF at a certain spatial frequency. From a signal processing
    > perspective it is sort of a low pass filter, which attenuates the
    > high frequencies.


    ... and the sensor and anti-alias filters have an MTF as well, as do the
    other elements of the system. Using a single value like "resolution" is
    just an approximation, that may, or may not, work well.

    Cheers,
    David
    David J Taylor, Feb 18, 2009
    #4
  5. Scott W

    RichA Guest

    On Feb 17, 11:15 pm, Alan Browne <>
    wrote:
    > David J. Littleboy wrote:
    > > "Alan Browne" <> wrote:
    > >> Scott W wrote:

    >
    > >> Yes - more brain idle, though I would have wrote it:

    >
    > >>         1/res_out^2 = 1/res_lens^2 + 1/res_sens^2.

    >
    > >>> Still you are correct that the lens needs to be taken in to account,
    > >>> and can be the limiting factor.
    > >> Everything in series has a limiting factor.

    >
    > > FWIW, I find that it's quite possible to get corner-to-corner sharp images
    > > from the 5D2. Most of the wide angle lenses I own need to be stopped down to
    > > f/11 or f/16 (which is an irritation and is why I'll be buying the ZE 21/2.8
    > > if it comes out), but at 35mm and longer most lenses are sharp at the
    > > corners at f/5.6 or f/8.

    >
    > > So the lens is only the limiting factor if one isn't thinking/trying.

    >
    > Well, there is the difference between absolute performance, pixel
    > peeping and plain making good photography.
    >
    > Oddly, when I posted the "Megapixel challenge" before Christmas there
    > was only one taker (other than me) - and he shot on film (acquitting
    > himself well at that - I scanned his film on my 9000ED).  The photos
    > were corner shots only, BTW, though shot in the sweet aperture spot.


    Waste of time. I shot Kodak's new Ektar 100 and scanned it and it has
    about 1/2 the resolution of a D300 so your Sony A900 should wipe the
    floor with any 35mm film scan. Michael Reichman is right, film's day
    is done in this respect.
    RichA, Feb 19, 2009
    #5
  6. Scott W

    RichA Guest

    On Feb 18, 2:26 pm, "David J Taylor" <-
    this-bit.nor-this.co.uk> wrote:
    > Alfred Molon wrote:
    > > In article <RsRml.36722$>, David J
    > > Taylor says...

    >
    > >> It makes sense if you use the same spatial units, whether that be
    > >> microns or megapixels, and then  use the appropriate formula.  But
    > >> you are right in that using one figure for lens performance may
    > >> simplify things too much.

    >
    > > The problem is that a lens does not have a resolution - it has a
    > > certain MTF at a certain spatial frequency. From a signal processing
    > > perspective it is sort of a low pass filter, which attenuates the
    > > high frequencies.

    >
    > .. and the sensor and anti-alias filters have an MTF as well, as do the
    > other elements of the system.  Using a single value like "resolution" is
    > just an approximation, that may, or may not, work well.
    >
    > Cheers,
    > David


    Not to mention that the subject has an effect on resolution. If you
    shoot 3" wide power lines against a sky at a distance of 5 miles,
    you'll see them on the sensor, even if their physical dimension makes
    theoretical resolution impossible. This is a case of contrast-induced
    "super resolution." Where as a low contrast subject of the same size
    will never show up on the sensor.
    RichA, Feb 19, 2009
    #6
  7. RichA wrote:
    []
    > Not to mention that the subject has an effect on resolution. If you
    > shoot 3" wide power lines against a sky at a distance of 5 miles,
    > you'll see them on the sensor, even if their physical dimension makes
    > theoretical resolution impossible. This is a case of contrast-induced
    > "super resolution." Where as a low contrast subject of the same size
    > will never show up on the sensor.


    The subject has zero effect on the system MTF, of course.

    However, the subject contrast will determine how high a spatial frequency
    you can use before the signal from the subect is lost in the system noise.
    We used to use the term "minimum resolvable modulation", but "minimum
    resolvable contrast" may be more in fashion now.

    http://en.wikipedia.org/wiki/Minimum_resolvable_contrast

    When you have a line (or point) source, what I think you are saying is
    that 20% modulation over one pixel may effectively detected in the eye as
    5% modulation over 4 pixels - for example.

    Cheers,
    David
    David J Taylor, Feb 19, 2009
    #7
  8. Alfred Molon wrote:
    > In article <RhZml.36934$>, David J
    > Taylor says...
    >> Alfred Molon wrote:
    >>> In article <RsRml.36722$>, David
    >>> J Taylor says...
    >>>
    >>>> It makes sense if you use the same spatial units, whether that be
    >>>> microns or megapixels, and then use the appropriate formula. But
    >>>> you are right in that using one figure for lens performance may
    >>>> simplify things too much.
    >>>
    >>> The problem is that a lens does not have a resolution - it has a
    >>> certain MTF at a certain spatial frequency. From a signal processing
    >>> perspective it is sort of a low pass filter, which attenuates the
    >>> high frequencies.

    >>
    >> .. and the sensor and anti-alias filters have an MTF as well, as do
    >> the other elements of the system. Using a single value like
    >> "resolution" is just an approximation, that may, or may not, work
    >> well.

    >
    > The sensor has a pixel count - something the filter and the lens do
    > not. You're mixing up things.


    The sensor, filter and lens all have an MTF, as does the Bayer processing.
    All the same - the transfer function at a certain cycles per mm (or
    whatever unit you choose). Of course, you need to measure the MTF in the
    same spatial units, but it doesn't matter what those spatial units are.

    David
    David J Taylor, Feb 19, 2009
    #8
  9. Alfred Molon wrote:
    > In article <bghnl.37369$>, David J
    > Taylor says...
    >
    >> The sensor, filter and lens all have an MTF, as does the Bayer
    >> processing. All the same - the transfer function at a certain cycles
    >> per mm (or whatever unit you choose). Of course, you need to
    >> measure the MTF in the same spatial units, but it doesn't matter
    >> what those spatial units are.

    >
    > The sensor is a discrete sampler and for practical purposes you
    > consider it the source of the signal. All other things, the AA
    > filter, the lens and Bayer interpolation have an impact on the high
    > frequency content.


    If you want to know the overall performance of the system, all the
    components must be taken into account. Changes in the MTF before and
    after the sensor can have distinctly different effects on the image,
    though. I'm thinking aliasing here.

    Cheers,
    David
    David J Taylor, Feb 19, 2009
    #9
    1. Advertising

Want to reply to this thread or ask your own question?

It takes just 2 minutes to sign up (and it's free!). Just click the sign up button to choose a username and then you can ask your own questions on the forum.
Similar Threads
  1. Mark Johnson

    Substantial savings?

    Mark Johnson, Feb 20, 2004, in forum: Digital Photography
    Replies:
    1
    Views:
    275
  2. Bill Hilton

    39 megapixels? 31 megapixels? Get 'em here ...

    Bill Hilton, Jul 16, 2005, in forum: Digital Photography
    Replies:
    7
    Views:
    330
    Bill Hilton
    Jul 18, 2005
  3. hillyer
    Replies:
    0
    Views:
    418
    hillyer
    Jan 30, 2007
  4. Peter Potamus the Purple Hippo

    Re: Mozilla versus IE versus Opera versus Safari

    Peter Potamus the Purple Hippo, May 8, 2008, in forum: Firefox
    Replies:
    0
    Views:
    675
    Peter Potamus the Purple Hippo
    May 8, 2008
  5. jdanield
    Replies:
    0
    Views:
    248
    jdanield
    Oct 17, 2012
Loading...

Share This Page