Pentax K-3 : the first dSLR with selectable antialias filter

Discussion in 'Digital Cameras' started by Nick Fotis, Oct 12, 2013.

  1. Nick Fotis

    Nick Fotis Guest

    http://www.dpreview.com/previews/pentax-k-3/

    Looks interesting as a concept.

    And not being forced to pay 300 Eurodollars for the priviledge of having
    the AA filter off sounds nice.

    Still wondering why they didn't make a FF body, yet.

    N.F.
     
    Nick Fotis, Oct 12, 2013
    #1
    1. Advertisements

  2. Nick Fotis

    Alfred Molon Guest

    If it works it's a good idea. But I read that this only works when the exposure
    time is longer than 1/1000s.
     
    Alfred Molon, Oct 12, 2013
    #2
    1. Advertisements

  3. Nick Fotis

    Alan Browne Guest

    That would cover over 99% of my shooting.


    --
    "Political correctness is a doctrine, fostered by a delusional,
    illogical minority, and rapidly promoted by mainstream media,
    which holds forth the proposition that it is entirely possible
    to pick up a piece of shit by the clean end."
    -Unknown
     
    Alan Browne, Oct 12, 2013
    #3
  4. Nick Fotis

    Alan Browne Guest

    If I read it correctly it seems to be a non-AA sensor that gets
    mechanically nudged during exposure to soften the image.

    QUOTE Pentax is using its sensor-shift image stabilizer to deliberately
    move the sensor during the exposure, slightly blurring the image to
    mimic the effects of an optical low-pass filter. ENDQUOTE.

    If that's the case then it would seem to me that you could do that
    effect in post processing.

    --
    "Political correctness is a doctrine, fostered by a delusional,
    illogical minority, and rapidly promoted by mainstream media,
    which holds forth the proposition that it is entirely possible
    to pick up a piece of shit by the clean end."
    -Unknown
     
    Alan Browne, Oct 12, 2013
    #4
  5. Nick Fotis

    RichA Guest

    Do NOT deny Nikon the ability to make money!!!!
     
    RichA, Oct 12, 2013
    #5
  6. Nick Fotis

    Alfred Molon Guest

    No you can't. The low pass filter must be applied *before* the sampling (see
    Nyqquist's theorem).
     
    Alfred Molon, Oct 12, 2013
    #6
  7. Nick Fotis

    Alan Browne Guest

    I'm aware of that - and I've made the mistake in the past of believing
    that AA could be applied in post.

    However, here the "blur" is made by moving the sensor ("the sampler") at
    the time of sampling so it's not clear to me that that is any different
    than doing it in post since you can simulate the same thing.

    One has to accept that moving the sample while sampling (a means to blur
    no different than vibration or camera shake) to smooth out aliasing.
    Blur is something you can do in post many ways.

    --
    "Political correctness is a doctrine, fostered by a delusional,
    illogical minority, and rapidly promoted by mainstream media,
    which holds forth the proposition that it is entirely possible
    to pick up a piece of shit by the clean end."
    -Unknown
     
    Alan Browne, Oct 12, 2013
    #7
  8. Nick Fotis

    Alfred Molon Guest

    The motion blur is applied before the sampling, i.e. the sensor captures a
    slightly blurred image.
    Think of it as the sensor being still and the world around the camera slightly
    moving at high speed.
    The end effect is that the sensor sees a slightly blurred image.
     
    Alfred Molon, Oct 13, 2013
    #8
  9. Nick Fotis

    Alan Browne Guest

    Not at all - the blur is applied _while_ sampling.
    Yes. When I wrote: " However, here the "blur" is made by moving the
    sensor ("the sampler") at the time of sampling "

    that I was clear on how the image was blurred. That that constituted
    pre-sampling filtering is what I am asking. It is not pre-sampling it
    is the sampler - can you AA at the sampler? Should it not be done
    before the sampler?

    The question remains: is this "mimic" of an AA just a way to blur the
    image that could just as easily be simulated in post processing?



    --
    "Political correctness is a doctrine, fostered by a delusional,
    illogical minority, and rapidly promoted by mainstream media,
    which holds forth the proposition that it is entirely possible
    to pick up a piece of shit by the clean end."
    -Unknown
     
    Alan Browne, Oct 13, 2013
    #9
  10. Nick Fotis

    Alfred Molon Guest

    Let's rephrase it then:
    1. The sensor sees a blurred image.
    2. The sensor samples this blurred image.

    The blur is caused by the rapid motion of the sensor, which has the same effect
    as a low pass filter.
    Of course not. The low pass filter has to be applied before the sampling.
     
    Alfred Molon, Oct 14, 2013
    #10
  11. Nick Fotis

    Alan Browne Guest

    I understand what you're saying. I understood it the first time you
    said it, and the 2nd and the third. I am long past that.

    What I dispute is if it is somehow, really, equivalent to the effect of
    AA or whether it is simply another way of blurring an image - something
    most easily done in post processing.


    --
    "Political correctness is a doctrine, fostered by a delusional,
    illogical minority, and rapidly promoted by mainstream media,
    which holds forth the proposition that it is entirely possible
    to pick up a piece of shit by the clean end."
    -Unknown
     
    Alan Browne, Oct 14, 2013
    #11
  12. Nick Fotis

    Joe Kotroczo Guest

    On 12/10/2013 16:14, Nick Fotis wrote:

    (...)
    Losing a lot of FF lenses from their lens line-up in recent years?
     
    Joe Kotroczo, Oct 14, 2013
    #12
  13. Nick Fotis

    Joe Kotroczo Guest

    On 14/10/2013 18:23, Alan Browne wrote:
    (...)
    From the article: "microscopic vibrations to the image sensor unit at
    the sub-pixel level during exposure'"

    Smallest unit you can manipulate in post is the pixel, no?
     
    Joe Kotroczo, Oct 14, 2013
    #13
  14. Nick Fotis

    Alan Browne Guest

    Very good point - but I see the applied method as simply forcing a mix
    of some information from the adjacent pixel to become present in the
    nominal pixel location. (And there's the RGB mosaic to consider as well
    I suppose).

    Further, a lot of post processing uses (interpolate) portions of
    adjacent pixels in computing an effect. eg: the Unsharp Mask,
    enlargements, possibly some size reduction algorithms.


    --
    "Political correctness is a doctrine, fostered by a delusional,
    illogical minority, and rapidly promoted by mainstream media,
    which holds forth the proposition that it is entirely possible
    to pick up a piece of shit by the clean end."
    -Unknown
     
    Alan Browne, Oct 14, 2013
    #14
  15. His above statement answers your below question.

    It *cannot* be simulated in post processing done after
    sampling for the simple reason that sampling changes the
    frequency domain of the image data. The higher
    frequency components that are more affected by any kind
    of a low pass filter (i.e., blurring effect) are no
    longer at the same higher frequencies and are no longer
    more affected.

    Pre-sampling data is not the same data as is available
    post-sampling. Pre-filtering does not have the same
    effect as post-filtering.
    An anti-aliasing filter *is* a "way of blurring an
    image". Low pass filter, AA filter, blur... are all the
    same thing. The affect on an image before sampling is
    distinctly different than the effect on an image after
    sampling.

    The effect of "Anti-aliasing" filtering (or call it low
    pass filtering or call it blurring) of pre-sampled data
    cannot be achieved after sampling because the data is
    fundamentally changed by the sampling process.
     
    Floyd L. Davidson, Oct 14, 2013
    #15
  16. Nick Fotis

    Eric Stevens Guest

    Actually, during the sampling.
     
    Eric Stevens, Oct 14, 2013
    #16
  17. It necessarily has to be *before* sampling. The sample has to be affected
    by the blur, which must have already taken place.
     
    Floyd L. Davidson, Oct 14, 2013
    #17
  18. Nick Fotis

    Alan Browne Guest

    So would you call the method Pentax use "AA" ish, mimic-ish or something
    that could be applied in post?


    --
    "Political correctness is a doctrine, fostered by a delusional,
    illogical minority, and rapidly promoted by mainstream media,
    which holds forth the proposition that it is entirely possible
    to pick up a piece of shit by the clean end."
    -Unknown
     
    Alan Browne, Oct 14, 2013
    #18
  19. Nick Fotis

    Alfred Molon Guest

    The question is to what extent the low pass filter effect of shaking quickly
    the sensor corresponds to a real AA filter.

    The ideal AA filter has a step response, i.e. cuts off all frequencies beyond
    Nyquist, and nothing below.

    The Pentax K3 sensor shake low pass filtering will differ a bit and I suspect
    not be as good as a real AA filter (which in turn is not ideal).

    You'll probably get some degree of AA filtering, which probably is better than
    nothing. I think we need to wait for the reviews to get an idea how this sensor
    shake AA filter works.
     
    Alfred Molon, Oct 14, 2013
    #19
  20. Don't be so skeptical! Think about the potential...

    First, consider that the sensor travel has to be fast
    enough that at least 1 complete cycle must be made for
    each and every pixel while the shutter is open for that
    pixel. With a focal plane shutter that requires a
    "shake frequency" multiple times the fastest shutter
    speed. And while it needs at least 1 cycle, if it only
    makes 1 cycle the motion necessarily must be very smooth
    across the entire exposure time. But if it can do, say,
    100 cycles for each pixel the smoothness of movement
    required is greatly reduced. Which is to say that if it
    is fast enough the shape factor could be very nice, with
    very few artifacts introduced.

    We can assume a circular or eliptical movement, though
    going through something like a square cycle would have
    the same effect.

    Now consider the potential: the "degree of AA filtering"
    would be determined by the amount of sensor travel, and
    by varying the size of the circle the sensor is moved
    through, the AA filter can be have a variable cutoff
    frequency.

    A variable AA filter. And one that has a much sharper
    cutoff than any existing optical filter.
     
    Floyd L. Davidson, Oct 14, 2013
    #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.