Sign in with
Sign up | Sign in
Your question

Sharpening artefacts and MTF of monitors?

Last response: in Digital Camera
Share
Anonymous
a b C Monitor
April 20, 2005 5:05:34 AM

Archived from groups: rec.photo.digital (More info?)

I'm trying to understand visible effects of artefacts of sharpening with
different viewing conditions. Suppose I have a sharpening matrix such
that I can't see artefacts on monitor (with a very good CRT monitor,
VS P815). However, when I blow pixels 200%, a careful examination
exposes some artefacts. With magnification about 500% artefacts become
very visible.

Given this data, can I estimate how the image will be seen on LCD monitors
(which have better MTF at high spacial frequencies)? Can I estimate
on which pixel-per-inch value the printed image will have visible
artefacts?

In short: what makes the artefact invisible without magnification: bad
MTF of the CRT monitor (due to internal reflection in glass), or bad MTF
of the eye? If the latter, then a print with about 120 pixels/inch or
more should show no artefacts (pixel size on the monitor is 0.26mm).
If the former, then one needs much higher pixels/inch value...

================

Here is an example: take the image

http://www.dpreview.com/reviews/KonicaMinoltaA200/sampl...

This is a resolution chart, the artefacts should be very visible. I run
DSP software, and it shows the following MTF curve of the combined
lens/sensor/demosaicer:

1 """"xxx__''''''''''''''''''''''''''''''''''''''''''''''''''''''|
| ""x_ |
| "x_ |
| "x_ |
| "x_ |
| "_ |
| "x_ |
| x_ |
| "x |
| "x_ |
| "_ |
| "x_ |
| "_ | |
| "x | |
| "x_ | |
| x_ | |
| "_ | |
| "x |
| |"x_ |
| | x_ |
| | "x_ |
-0.016------------------------------------------------------------"xx_
0 3000

(vertical line is the Nyquist frequency of the sensor). One can easily
see that sharpening with the matrix

0 -0.375 0
-0.375 2.5 -0.375
0 -0.375 0

makes the throughput MTF curve almost horizontal:


1 """"""""""""""""xxxx____'''''''''''''''''''''''''''''''''''''''|
| ""xx__ |
| ""xx_ |
| "xx_ |
| "_ |
| "x |
| "x |
| "_ |
| x |
| "_ | |
| _| |
| x |
| |x |
| | x |
| | x |
| | " |
| | " |
| | " |
| | "_ |
| | _ |
,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,|,,,,,,,,,x,,
-.05|..................................................|.........."_
0 3000

Indeed, this sharpening has enormous "clearing" effect on the picture;
the artefacts are as described above: invisible on CRT without
magnification (initial image has no artefacts). It would be
interesting to know what happens with other media...

Thanks,
Ilya
Anonymous
a b C Monitor
April 20, 2005 5:05:35 AM

Archived from groups: rec.photo.digital (More info?)

In article <d449su$1aih$1@agate.berkeley.edu>,
Ilya Zakharevich <nospam-abuse@ilyaz.org> wrote:

> I'm trying to understand visible effects of artefacts of sharpening with
> different viewing conditions. Suppose I have a sharpening matrix such
> that I can't see artefacts on monitor (with a very good CRT monitor,
> VS P815). However, when I blow pixels 200%, a careful examination
> exposes some artefacts. With magnification about 500% artefacts become
> very visible.
>
> Given this data, can I estimate how the image will be seen on LCD monitors
> (which have better MTF at high spacial frequencies)? Can I estimate
> on which pixel-per-inch value the printed image will have visible
> artefacts?
>
> In short: what makes the artefact invisible without magnification: bad
> MTF of the CRT monitor (due to internal reflection in glass), or bad MTF
> of the eye? If the latter, then a print with about 120 pixels/inch or
> more should show no artefacts (pixel size on the monitor is 0.26mm).
> If the former, then one needs much higher pixels/inch value...
>
> ================
>
> Here is an example: take the image
>
> http://www.dpreview.com/reviews/KonicaMinoltaA200/sampl...
>
> This is a resolution chart, the artefacts should be very visible. I run
> DSP software, and it shows the following MTF curve of the combined
> lens/sensor/demosaicer:
>
> 1 """"xxx__''''''''''''''''''''''''''''''''''''''''''''''''''''''|
> | ""x_ |
> | "x_ |
> | "x_ |
> | "x_ |
> | "_ |
> | "x_ |
> | x_ |
> | "x |
> | "x_ |
> | "_ |
> | "x_ |
> | "_ | |
> | "x | |
> | "x_ | |
> | x_ | |
> | "_ | |
> | "x |
> | |"x_ |
> | | x_ |
> | | "x_ |
> -0.016------------------------------------------------------------"xx_
> 0 3000
>
> (vertical line is the Nyquist frequency of the sensor). One can easily
> see that sharpening with the matrix
>
> 0 -0.375 0
> -0.375 2.5 -0.375
> 0 -0.375 0
>
> makes the throughput MTF curve almost horizontal:
>
>
> 1 """"""""""""""""xxxx____'''''''''''''''''''''''''''''''''''''''|
> | ""xx__ |
> | ""xx_ |
> | "xx_ |
> | "_ |
> | "x |
> | "x |
> | "_ |
> | x |
> | "_ | |
> | _| |
> | x |
> | |x |
> | | x |
> | | x |
> | | " |
> | | " |
> | | " |
> | | "_ |
> | | _ |
> ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,|,,,,,,,,,x,,
> -.05|..................................................|.........."_
> 0 3000
>
> Indeed, this sharpening has enormous "clearing" effect on the picture;
> the artefacts are as described above: invisible on CRT without
> magnification (initial image has no artefacts). It would be
> interesting to know what happens with other media...
>
> Thanks,
> Ilya




Sharpening could work in this case. In some cases you could see ripples
in that chart or the S/N ratio will get bad.
Anonymous
a b C Monitor
April 20, 2005 10:37:38 AM

Archived from groups: rec.photo.digital (More info?)

[A complimentary Cc of this posting was sent to
Kevin McMurtrie
<mcmurtri@dslextreme.com>], who wrote in article <mcmurtri-2C7294.21195119042005@corp-radius.supernews.com>:

> > 1 """"""""""""""""xxxx____'''''''''''''''''''''''''''''''''''''''|
> > | ""xx__ |
> > | ""xx_ |
> > | "xx_ |
> > | "_ |
> > | "x |
> > | "x |
> > | "_ |
> > | x |
> > | "_ | |
> > | _| |
> > | x |
> > | |x |
> > | | x |
> > | | x |
> > | | " |
> > | | " |
> > | | " |
> > | | "_ |
> > | | _ |
> > ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,|,,,,,,,,,x,,
> > -.05|..................................................|.........."_
> > 0 3000

> Sharpening could work in this case. In some cases you could see ripples
> in that chart or the S/N ratio will get bad.

Of course, S/N is decreased about 2.6 times by this sharpening. So it
depends on what is the initial value of S/N. I did not measure S/N of
the Adobe demosaicer; with the builtin demozaicer of the camera it is
about 60 (for luma of 18% gray, at ISO50). Probably such a good RTF
curve should result in larger noise than the lousy MTF curve of the
builtin demozaicer... Still, it may happen that S/N is above 20
(i.e., tolerable). Anyway, it is going to be a usual tradeoff between
detail and noise; note that one can process different parts of the
image with different settings.

About ripples: with MTF curve such as above, I doubt that any artefact
like "ripples" (whatever it is ;-) may be possible; there is a light
halo about dark areas (visible only under ). However, I was
seriously surprised about the quality of the image; having MTF above
50% corner-to-corner for an equivalent of 38 lp/mm (for 35mm film)
without visible artefacts gives very counterintuitive results...

Yours,
Ilya
Related resources
Can't find your answer ? Ask !
Anonymous
a b C Monitor
April 20, 2005 11:16:18 AM

Archived from groups: rec.photo.digital (More info?)

I wrote in article <d449su$1aih$1@agate.berkeley.edu>:

Throughput MTF curve of lens+sensor+demosaicer+sharpening is

> 1 """"""""""""""""xxxx____'''''''''''''''''''''''''''''''''''''''|
> | ""xx__ |
> | ""xx_ |
> | "xx_ |
> | "_ |
> | "x |
> | "x |
> | "_ |
> | x |
> | "_ | |
> | _| |
> | x |
> | |x |
> | | x |
> | | x |
> | | " |
> | | " |
> | | " |
> | | "_ |
> | | _ |
> ,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,,|,,,,,,,,,x,,
> -.05|..................................................|.........."_
> 0 3000
>
> Indeed, this sharpening has enormous "clearing" effect on the picture;
> the artefacts are as described above: invisible on CRT without
> magnification (initial image has no artefacts). It would be
> interesting to know what happens with other media...

BTW, I temporarily put the sharpened picture on
http://ilyaz.org/software/tmp/KM_A200-resolution-chart-...

It is 1.3M file (saved with "85% quality"; should have I chosen a
better quality?).

Yours,
Ilya

P.S. I calculated MTF to the cut-off frequency of the lens (f/4):

1 """"""xx_''''''''''''''''''''''''''''''''''''''''''''''''''''''|
| "x_ |
| "_ |
| x |
| x |
| x |
| _ | |
| | |
| x| |
| | |
| x |
| : |
| |: |
| |x |
| |: |
| | : |
| | " |
| | |
| | " __ _
0 -----------------------x-----__xxxxxxxxxxx_-----------x"--"xxx"-
| | "__x" "x_ _" |
-.123|............................................."x___"...........|
0 8000

(the vertical line is at Nyquist of the sensor, anything to the right
results in aliasing; but the data too far to the right is probably not
very reliable; I doubt that the second "dip" in the graph is actually
there). It is increadibly good; very little aliasing and very high
MTF in the main area; anyone having an idea how Adobe does it?
Anonymous
a b C Monitor
April 21, 2005 1:02:46 AM

Archived from groups: rec.photo.digital (More info?)

[A complimentary Cc of this posting was sent to
Kevin McMurtrie
<mcmurtri@dslextreme.com>], who wrote in article <mcmurtri-2C7294.21195119042005@corp-radius.supernews.com>:

> Sharpening could work in this case. In some cases you could see ripples
> in that chart or the S/N ratio will get bad.

I measured the S/N ratio after this sharpening. It looks like S/N of
18% gray should be 36. This is the S/N ratio of luma; thus S/N ratio
of luminosity is 16; so it coincides with S/N ratio of Velvia 50 (as
given on Roger Clark's web site).

Yours,
Ilya
Anonymous
a b C Monitor
April 21, 2005 2:07:11 PM

Archived from groups: rec.photo.digital (More info?)

Ilya Zakharevich wrote:
> [A complimentary Cc of this posting was sent to
> Kevin McMurtrie
> <mcmurtri@dslextreme.com>], who wrote in article
> <mcmurtri-2C7294.21195119042005@corp-radius.supernews.com>:
>
>> Sharpening could work in this case. In some cases you could see
>> ripples in that chart or the S/N ratio will get bad.
>
> I measured the S/N ratio after this sharpening. It looks like S/N of
> 18% gray should be 36. This is the S/N ratio of luma; thus S/N ratio
> of luminosity is 16; so it coincides with S/N ratio of Velvia 50 (as
> given on Roger Clark's web site).
>
> Yours,
> Ilya

But the broadband signal-to-noise ratio may not reflect how the image is
perceived. You need to know the narrowband SNR, and weight that with the
visual systems frequency response (as is done with audio and weighted SNR
measurements). This will be particularly important in comparing different
media (e.g. film and digital).

David
Anonymous
a b C Monitor
April 22, 2005 12:56:37 AM

Archived from groups: rec.photo.digital (More info?)

[A complimentary Cc of this posting was sent to
David J Taylor
<david-taylor@blueyonder.co.not-this-bit.nor-this-part.uk>], who wrote in article <jXK9e.16662$G8.13913@text.news.blueyonder.co.uk>:
> > I measured the S/N ratio after this sharpening. It looks like S/N of
> > 18% gray should be 36. This is the S/N ratio of luma; thus S/N ratio
> > of luminosity is 16; so it coincides with S/N ratio of Velvia 50 (as
> > given on Roger Clark's web site).

> But the broadband signal-to-noise ratio may not reflect how the image is
> perceived. You need to know the narrowband SNR, and weight that with the
> visual systems frequency response (as is done with audio and weighted SNR
> measurements).

Thanks, this may be relevant in some other situation. But given that
MRF of the whole workflow is provided in the initial message, and is
"almost horizontal", weighting will not change things much on the
digital side.

On the film side - I know no data about narrow band SNR of Velvia 50;
do you? And the MTF curve of Velvia 50 + reasonably good fixed focal
length lens is going to be "similarly good" when you compensate for 4x
difference in sensor size, so the broadband S/N should be viable
too...

Anyway, it may be that your suggestions will change the numbers about
25%; but I expect the change to be in the same direction for digital
and film. Given that I suspect very much
  • the data for film noise
    on Roger's site, such a correction is not very important.

    Thanks,
    Ilya

  • E.g., his S/N=16 for 18% gray on Velvia 50

    http://clarkvision.com/imagedetail/digital.signal.to.no...

    does not specify the window size into the film; he compares with a
    camera which has 8M pixels, and 8.2 micron cells, so there may be
    different interpretations of "equivalent" window size into the film.
    In a private communication he says that the window size is actually
    6.3 microns (equivalent of 24M pixels).

    Actually, comparing noise of 24MP scan with noise of 8MP digital
    sensor may be not that crazy, given his "translation rules" from scan
    resolution to digital sensor resolution (in other papers on his site).
    However, it is clear that one can reduce noise of 24MP scan by
    post-processing without lowering the resolution much (comparing to 8MP
    scan); I do not know whether this is taken into account.

    Additionally, it is not specified whether S/N=16 is for noise of
    density, or noise of "initial luminance". Given high contrast of
    slide film (gamma = 1/1.5), the latter should be about 66% of the
    former... Roger still did not answer my email about this issue.
    Anonymous
    a b C Monitor
    April 22, 2005 12:45:10 PM

    Archived from groups: rec.photo.digital (More info?)

    Ilya Zakharevich wrote:
    > [A complimentary Cc of this posting was sent to
    > David J Taylor
    > <david-taylor@blueyonder.co.not-this-bit.nor-this-part.uk>], who
    > wrote in article <jXK9e.16662$G8.13913@text.news.blueyonder.co.uk>:
    >>> I measured the S/N ratio after this sharpening. It looks like S/N
    >>> of 18% gray should be 36. This is the S/N ratio of luma; thus S/N
    >>> ratio of luminosity is 16; so it coincides with S/N ratio of Velvia
    >>> 50 (as given on Roger Clark's web site).
    >
    >> But the broadband signal-to-noise ratio may not reflect how the
    >> image is perceived. You need to know the narrowband SNR, and weight
    >> that with the visual systems frequency response (as is done with
    >> audio and weighted SNR measurements).
    >
    > Thanks, this may be relevant in some other situation. But given that
    > MRF of the whole workflow is provided in the initial message, and is
    > "almost horizontal", weighting will not change things much on the
    > digital side.

    I'm not so sure about "almost horizontal MTF". For any sampled-data
    system there needs to be an antialias filter present prior to the sampling
    to avoid artefacts. The frequency response shape of this filter varies
    considerably between cameras (e.g. some cameras have none just to boost
    the system high-frequency response). Depending on the distance of the eye
    from the viewed image, the maximum sensitivity of the eye might fall in
    this rather variable MTF range.

    > On the film side - I know no data about narrow band SNR of Velvia 50;
    > do you? And the MTF curve of Velvia 50 + reasonably good fixed focal
    > length lens is going to be "similarly good" when you compensate for 4x
    > difference in sensor size, so the broadband S/N should be viable
    > too...
    >
    > Anyway, it may be that your suggestions will change the numbers about
    > 25%; but I expect the change to be in the same direction for digital
    > and film. Given that I suspect very much
  • the data for film noise
    > on Roger's site, such a correction is not very important.
    >
    > Thanks,
    > Ilya

    No, I have no experience of measuring film in this way, my experience is
    entirely with electronic imaging systems (visible, thermal etc.). I do
    suspect that the different amplitude linearity, MTF and noise sprecta is
    what distinguishes "digital" from "film", and that some people will prefer
    one to the other. Perhaps a lifetime of looking at film images makes the
    brian adjust to thinking them "normal".

    Cheers,
    David
    Anonymous
    a b C Monitor
    April 22, 2005 1:10:01 PM

    Archived from groups: rec.photo.digital (More info?)

    [A complimentary Cc of this posting was sent to
    David J Taylor
    <david-taylor@blueyonder.co.not-this-bit.nor-this-part.uk>], who wrote in article <qQ2ae.17300$G8.3943@text.news.blueyonder.co.uk>:
    > > Thanks, this may be relevant in some other situation. But given that
    > > MRF of the whole workflow is provided in the initial message, and is
    > > "almost horizontal", weighting will not change things much on the
    > > digital side.

    > I'm not so sure about "almost horizontal MTF".

    Did not you see the curve in the initial posting? Or are you using
    proportional width fonts in your Usenet reading (which is a no-no-no
    ;-)? Then you need to cut&paste the graph to an editor with
    monospaced font (e.g. Courier).

    > For any sampled-data system there needs to be an antialias filter
    > present prior to the sampling to avoid artefacts.

    Any photo camera has such a filter builtin; it is called a lens. ;-)
    Of course, for interchangeable-lens cameras one needs a way to
    compensate for difference in lens quality. (This is probably the
    reason why dSLR have such pitiful MTFs...)

    > No, I have no experience of measuring film in this way, my experience is
    > entirely with electronic imaging systems (visible, thermal etc.). I do
    > suspect that the different amplitude linearity, MTF and noise sprecta is
    > what distinguishes "digital" from "film", and that some people will prefer
    > one to the other. Perhaps a lifetime of looking at film images makes the
    > brian adjust to thinking them "normal".

    Hmm, a very interesting idea. I need to think about it more...

    Thanks,
    Ilya
    Anonymous
    a b C Monitor
    April 22, 2005 1:21:45 PM

    Archived from groups: rec.photo.digital (More info?)

    Ilya Zakharevich wrote:
    []
    >> I'm not so sure about "almost horizontal MTF".
    >
    > Did not you see the curve in the initial posting? Or are you using
    > proportional width fonts in your Usenet reading (which is a no-no-no
    > ;-)? Then you need to cut&paste the graph to an editor with
    > monospaced font (e.g. Courier).

    No, I was responding about the MTF of typical camera equipment.

    >> For any sampled-data system there needs to be an antialias filter
    >> present prior to the sampling to avoid artefacts.
    >
    > Any photo camera has such a filter builtin; it is called a lens. ;-)
    > Of course, for interchangeable-lens cameras one needs a way to
    > compensate for difference in lens quality. (This is probably the
    > reason why dSLR have such pitiful MTFs...)

    Well, relying on the lens as the sole anti-alias filter is not adequate,
    at least not for interchangable lens cameras. [I have always felt that
    the Nikon 5700 produces such good results because its lens and sensor are
    "well-matched", probably that the lens is carefully designed not to excite
    artefacts in the sensor by keeping the MTF under control.] It's the
    "pitiful MTFs" coupled with the much better signal-to-noise ratio which
    produce a characteristically different image between film and digital.
    It's then subjective as to which is "better"!

    Cheers,
    David
    Anonymous
    a b C Monitor
    April 25, 2005 10:31:41 AM

    Archived from groups: rec.photo.digital (More info?)

    [A complimentary Cc of this posting was sent to
    David J Taylor
    <david-taylor@blueyonder.co.not-this-bit.nor-this-part.uk>], who wrote in article <Jm3ae.17315$G8.9433@text.news.blueyonder.co.uk>:

    > >> I'm not so sure about "almost horizontal MTF".

    > > Did not you see the curve in the initial posting? Or are you using
    > > proportional width fonts in your Usenet reading (which is a no-no-no
    > > ;-)? Then you need to cut&paste the graph to an editor with
    > > monospaced font (e.g. Courier).

    > No, I was responding about the MTF of typical camera equipment.

    Well, what I meant is the MTF of a particular pipeline through a
    particular camera. Actually, my initial calculations of "sharpening
    which compensates the MTF curve" were faulty (I used linear
    appriximation to trig functions while they were well accessible
    directly); the corrected result has following properties:

    a) MTF within 0.15% of 1 up to 0.4 of Nyquist frequency; 0.99% at
    0.5 of Nyquist; 0.95 at 0.6 of Nyquist; 75% at 80% of Nyquist (this
    is close to center; but the radial and tangential data at corner are
    very similar to that at center);

    b) No visual artefacts (for Gibbs phenomenon);

    c) Visual noise is slightly increased by this sharpening, but this
    is hardly noticable (the noise *is* increased at high frequency, but
    probably the eye is not sensible in this region);

    > > compensate for difference in lens quality. (This is probably the
    > > reason why dSLR have such pitiful MTFs...)

    > Well, relying on the lens as the sole anti-alias filter is not adequate,
    > at least not for interchangable lens cameras.

    Well, the fact that DSLR give *very* similar MTF with prime lens
    (presumably at f/8) as non-interchangeble lens cameras with 8x zooms
    at f/4 says enough....

    Thanks,
    Ilya
    Anonymous
    a b C Monitor
    April 29, 2005 12:46:05 PM

    Archived from groups: rec.photo.digital (More info?)

    [A complimentary Cc of this posting was NOT [per weedlist] sent to
    Ilya Zakharevich
    <nospam-abuse@ilyaz.org>], who wrote in article <d44vk2$1gqo$1@agate.berkeley.edu>:

    > BTW, I temporarily put the sharpened picture on
    > http://ilyaz.org/software/tmp/KM_A200-resolution-chart-...

    As I said in another message on this thread, the image above is
    oversharpened at low frequencies (due to an error on my part: I used a
    linear approximation at one point of calculations). Sorry. The image
    is now removed.

    A corrected image is on

    http://ilyaz.org/software/tmp/KM_A200-resolution-chart-...

    It is obtained in the following way:

    a) Take the Adobe-demosaiced image cited in the original page; it
    already has most features of MTF curve corrected;

    b) Apply this convolution matrix (divided by 16):

    # 0 0 0.6 0 0
    # 0 1.2 -7.12 1.2 0
    # 0.6 -7.12 37.28 -7.12 0.6
    # 0 1.2 -7.12 1.2 0
    # 0 0 0.6 0 0

    c) Correct the wrong (?) gamma (I think the Adobe image has gamma=3
    instead of 2.2; do not know why; I just compared it with
    in-camera produced JPEG).

    > P.S. I calculated MTF to the cut-off frequency of the lens (f/4):

    New data:

    1.016 """"""""""xx_''''''''''''''''''''''''''''''''''''''''''''''''''|
    | "_ |
    | x |
    | x |
    | _ |
    | |
    | x |
    | |
    | " | |
    | | |
    | "| |
    | | |
    | x |
    | | |
    | |x |
    | | |
    | | x |
    | | |
    | | " _|
    0 ,,,,,,,,,,,,,,,,,,,,,,|,,_,,,,,,,,__________x___,,,,,,,,,,,_"","
    | | x _x" "x_ _" |
    -0.108|....................|....."x""...................."x___"......|
    0 7350

    As usual, the vertical line is at Nyquist frequency of the sensor.

    Yours,
    Ilya
    Anonymous
    a b C Monitor
    May 2, 2005 5:00:02 AM

    Archived from groups: rec.photo.digital (More info?)

    [A complimentary Cc of this posting was NOT [per weedlist] sent to
    Ilya Zakharevich
    <nospam-abuse@ilyaz.org>], who wrote in article <d4ss8d$1eu7$1@agate.berkeley.edu>:
    > A corrected image is on
    >
    > http://ilyaz.org/software/tmp/KM_A200-resolution-chart-...
    >
    > It is obtained in the following way:

    .... and I forgot to write what I want from *you*. Can somebody look
    at the image on a good quality LCD monitor and check whether artefacts
    are visible without blowing up the pixels? (I mean with 1 image pixel
    mapped to 1 screen pixel.)

    The artefacts which I can see with high magnification are:

    a) slight white "halo" about dark areas;

    b) very slight "light interior" in dark strokes. E.g., take number
    1; divide the vertical line into 3 parts: left, center and right.
    The center is slightly lighter than the left and right parts.

    Maybe somebody can see it even on a CRT? (I can't on mine.)

    [I would like to ask about prints, but can't, since they cost. My
    LJ1100 prints without artefacts visible; but given that it is only
    600dpi, this is hardly telling about anything.]

    Thanks,
    Ilya
    Anonymous
    a b C Monitor
    May 2, 2005 12:54:13 PM

    Archived from groups: rec.photo.digital (More info?)

    Ilya Zakharevich wrote:
    []
    >>
    >> http://ilyaz.org/software/tmp/KM_A200-resolution-chart-...

    I see that Konica Minolta /still/ haven't sorted out the nasty artefacts
    on fine edges - e.g. the horizontal resolution chart from about 12 onwards
    near the centre of the image. This defect has been on the A1, A2 and
    A200. Don't they care about it?

    David
    Anonymous
    a b C Monitor
    May 2, 2005 2:23:50 PM

    Archived from groups: rec.photo.digital (More info?)

    [A complimentary Cc of this posting was sent to
    David J Taylor
    <david-taylor@blueyonder.co.not-this-bit.nor-this-part.uk>], who wrote in article <VUlde.23299$G8.21490@text.news.blueyonder.co.uk>:
    > >> http://ilyaz.org/software/tmp/KM_A200-resolution-chart-...

    > I see that Konica Minolta /still/ haven't sorted out the nasty artefacts
    > on fine edges - e.g. the horizontal resolution chart from about 12 onwards
    > near the centre of the image. This defect has been on the A1, A2 and
    > A200. Don't they care about it?

    It is hard to understand what you are talking about here.

    First of all, I do not see what KM has to do with this: there is no
    KM firmware/software in the digital pipeline. The raw image was
    demosaiced by Adobe, then postprocessed by me.

    Second, can you be more specific? "Horizontal resolution" means the
    chart with vertical lines, right? I can't see anything wrong with
    them; actually, same applies to horizonal lines.

    So: what should I pay attention to?

    Thanks,
    Ilya
    Anonymous
    a b C Monitor
    May 2, 2005 2:51:46 PM

    Archived from groups: rec.photo.digital (More info?)

    Ilya Zakharevich wrote:
    > [A complimentary Cc of this posting was sent to
    > David J Taylor
    > <david-taylor@blueyonder.co.not-this-bit.nor-this-part.uk>], who
    > wrote in article <VUlde.23299$G8.21490@text.news.blueyonder.co.uk>:
    >>>> http://ilyaz.org/software/tmp/KM_A200-resolution-chart-...
    >
    >> I see that Konica Minolta /still/ haven't sorted out the nasty
    >> artefacts on fine edges - e.g. the horizontal resolution chart from
    >> about 12 onwards near the centre of the image. This defect has been
    >> on the A1, A2 and A200. Don't they care about it?
    >
    > It is hard to understand what you are talking about here.
    >
    > First of all, I do not see what KM has to do with this: there is no
    > KM firmware/software in the digital pipeline. The raw image was
    > demosaiced by Adobe, then postprocessed by me.
    >
    > Second, can you be more specific? "Horizontal resolution" means the
    > chart with vertical lines, right? I can't see anything wrong with
    > them; actually, same applies to horizonal lines.
    >
    > So: what should I pay attention to?
    >
    > Thanks,
    > Ilya

    Well, if you can't see the defects, you have no need to worry about them.
    I am referring to the horizontal lines near the centre-right of the image
    labelled 18 (on the left) to 6 on the right. Just to the right of the
    central zone-plate spot. There is a "dirtiness" in the lines, leading to
    initial spots around the 18 resolution label. Perhaps there isn't enough
    anti-aliasing?

    It was picture defects like this which caused me to return my Minolta A2
    for a refund.

    David
    Anonymous
    a b C Monitor
    May 2, 2005 11:34:15 PM

    Archived from groups: rec.photo.digital (More info?)

    [A complimentary Cc of this posting was sent to
    David J Taylor
    <david-taylor@blueyonder.co.not-this-bit.nor-this-part.uk>], who wrote in article <6Dnde.23351$G8.15299@text.news.blueyonder.co.uk>:

    > Well, if you can't see the defects, you have no need to worry about them.

    This threads is exactly about things not seen which may appear in
    different situations/viewing conditions. So thanks for pointing out
    some of what you can see.

    > I am referring to the horizontal lines near the centre-right of the image
    > labelled 18 (on the left) to 6 on the right. Just to the right of the
    > central zone-plate spot. There is a "dirtiness" in the lines, leading to
    > initial spots around the 18 resolution label. Perhaps there isn't enough
    > anti-aliasing?

    Can you see something like this on vertical lines?

    I blew the image out 300% and do not see anything which I can consider
    a "defect". There is some modulation in horizontal direction in the
    "gray area" between white and black (e.g., on the 5th line from top
    over "12" mark). Or do you mean, e.g., the 3rd white line from top
    over "6"?

    Thanks,
    Ilya
    Anonymous
    a b C Monitor
    May 3, 2005 11:51:37 AM

    Archived from groups: rec.photo.digital (More info?)

    Ilya Zakharevich wrote:
    > [A complimentary Cc of this posting was sent to
    > David J Taylor
    []
    >> I am referring to the horizontal lines near the centre-right of the
    >> image labelled 18 (on the left) to 6 on the right. Just to the
    >> right of the central zone-plate spot. There is a "dirtiness" in the
    >> lines, leading to initial spots around the 18 resolution label.
    >> Perhaps there isn't enough anti-aliasing?
    >
    > Can you see something like this on vertical lines?
    >
    > I blew the image out 300% and do not see anything which I can consider
    > a "defect". There is some modulation in horizontal direction in the
    > "gray area" between white and black (e.g., on the 5th line from top
    > over "12" mark). Or do you mean, e.g., the 3rd white line from top
    > over "6"?
    >
    > Thanks,
    > Ilya


    http://math.berkeley.edu/~ilya/software/tmp/KM_A200-res...


    Yes, it's similar on the verttical lines, although perhaps not quite as
    intense. I'm viewing the image at 1:1 magnification in Internet Explorer.
    Measuring the pixel locations from top left, the artefacts are in the
    region (X:1800, Y:1215). It could be integer overflow or underflow.

    David

    David
    Anonymous
    a b C Monitor
    May 4, 2005 3:33:06 AM

    Archived from groups: rec.photo.digital (More info?)

    [A complimentary Cc of this posting was sent to
    David J Taylor
    <david-taylor@blueyonder.co.not-this-bit.nor-this-part.uk>], who wrote in article <d4Gde.23977$G8.19793@text.news.blueyonder.co.uk>:

    > Yes, it's similar on the verttical lines, although perhaps not quite as
    > intense. I'm viewing the image at 1:1 magnification in Internet Explorer.

    What is probably more important: what monitor? LCD/CRT, dot pitch?

    > Measuring the pixel locations from top left, the artefacts are in the
    > region (X:1800, Y:1215). It could be integer overflow or underflow.

    OK, I will try to investigate this region. ;-)

    Meanwhile, I'm pretty sure the original Adobe picture should have
    none; see

    http://www.dpreview.com/reviews/KonicaMinoltaA200/sampl...

    Can you see any defect (except the low contrast [and wrong gamma ;-]) there?

    I'm pretty sure that the artefacts must be added by the sharpening *I*
    did. The convolution matrix was calculated by me, but I do not have
    the source code for the software which applies it; on the other hand,
    I do not think it can do the convolution "wrong". So it must be due
    to the convolution matrix.

    Thanks,
    Ilya
    Anonymous
    a b C Monitor
    May 4, 2005 11:32:05 AM

    Archived from groups: rec.photo.digital (More info?)

    Ilya Zakharevich wrote:
    > [A complimentary Cc of this posting was sent to
    > David J Taylor
    > <david-taylor@blueyonder.co.not-this-bit.nor-this-part.uk>], who
    > wrote in article <d4Gde.23977$G8.19793@text.news.blueyonder.co.uk>:
    >
    >> Yes, it's similar on the verttical lines, although perhaps not quite
    >> as intense. I'm viewing the image at 1:1 magnification in Internet
    >> Explorer.
    >
    > What is probably more important: what monitor? LCD/CRT, dot pitch?
    >
    >> Measuring the pixel locations from top left, the artefacts are in the
    >> region (X:1800, Y:1215). It could be integer overflow or underflow.
    >
    > OK, I will try to investigate this region. ;-)
    >
    > Meanwhile, I'm pretty sure the original Adobe picture should have
    > none; see
    >
    >
    > http://www.dpreview.com/reviews/KonicaMinoltaA200/sampl...
    >
    > Can you see any defect (except the low contrast [and wrong gamma ;-])
    > there?
    >
    > I'm pretty sure that the artefacts must be added by the sharpening *I*
    > did. The convolution matrix was calculated by me, but I do not have
    > the source code for the software which applies it; on the other hand,
    > I do not think it can do the convolution "wrong". So it must be due
    > to the convolution matrix.
    >
    > Thanks,
    > Ilya

    I'm viewing on a 19 inch LCD at 1280 x 1024 resolution, but the artefacts
    show equally well if I open the image with Paint Shop Pro 9 and zoom
    around.

    I have noticed that the MTF of this monitor is much better than the 19
    inch CRT it replaced, resulting in images which were previously correct
    now appearing too sharp. My previous definition of "correct" was to have
    no perceptible overshoot on black/white transitions (well dark grey/light
    grey transitions - ones which didn't saturate).

    The original:
    http://www.dpreview.com/reviews/KonicaMinoltaA200/sampl...

    shows just the same artefacts, both when viewed with Internet Explorer,
    Windows XP Preview, IrfanView, or Paint Shop Pro 9. I can't comment on
    the gamma of the original as I don't know what the grey levels were (there
    is no calibrated greyscale that I can see).

    Cheers,
    David
    Anonymous
    a b C Monitor
    May 30, 2005 4:37:49 AM

    Archived from groups: rec.photo.digital (More info?)

    [A complimentary Cc of this posting was sent to
    David J Taylor
    <david-taylor@blueyonder.co.not-this-bit.nor-this-part.uk>], who wrote in article <VT_de.24782$G8.21108@text.news.blueyonder.co.uk>:

    [Posted with a big delay; my apologies; the thread on
    http://groups-beta.google.com/group/rec.photo.digital/b...
    ]

    > I see that Konica Minolta /still/ haven't sorted out the nasty
    > artefacts on fine edges - e.g. the horizontal resolution chart from
    > about 12 onwards near the centre of the image. This defect has been
    > on the A1, A2 and A200. Don't they care about it?


    > >> Measuring the pixel locations from top left, the artefacts are in the
    > >> region (X:1800, Y:1215). It could be integer overflow or underflow.
    > >
    > > OK, I will try to investigate this region. ;-)

    I put this region (magnified 10x) to

    ilyaz.org/software/tmp/KM_A200-resolution-chart-ACRraw-crop-artefacts-10x.png

    Do you mean (for example) the 4 gray pixels in the horizontal "white"
    line in the middle of the middle of the image?

    How I interpret these pixels: there is (effectively) an error in
    "calculation" of position of straight edges; the error may be as high
    as 0.25 pixels. I can hardly call this a defect... ;-)

    (It may be annoying indeed that these "errors" come in periodic
    groups; still, it is so much sub-pixel...)

    > I'm viewing on a 19 inch LCD at 1280 x 1024 resolution

    This is a LCD with 0.29mm dot pitch. On my CRT with 0.26mm dot pitch
    (actual dot pitch; the monitor's is 0.24mm, IIRC: VS P815) the MTF is
    so much worse, that I do not see these without magnification...

    Thanks,
    Ilya
    Anonymous
    a b C Monitor
    May 30, 2005 2:43:54 PM

    Archived from groups: rec.photo.digital (More info?)

    Ilya Zakharevich wrote:
    > [A complimentary Cc of this posting was sent to
    > David J Taylor
    > <david-taylor@blueyonder.co.not-this-bit.nor-this-part.uk>], who
    > wrote in article <VT_de.24782$G8.21108@text.news.blueyonder.co.uk>:
    >
    > [Posted with a big delay; my apologies; the thread on
    > http://groups-beta.google.com/group/rec.photo.digital/b...
    > ]
    >
    >> I see that Konica Minolta /still/ haven't sorted out the nasty
    >> artefacts on fine edges - e.g. the horizontal resolution chart from
    >> about 12 onwards near the centre of the image. This defect has been
    >> on the A1, A2 and A200. Don't they care about it?
    >
    >
    >>>> Measuring the pixel locations from top left, the artefacts are in
    >>>> the region (X:1800, Y:1215). It could be integer overflow or
    >>>> underflow.
    >>>
    >>> OK, I will try to investigate this region. ;-)
    >
    > I put this region (magnified 10x) to
    >
    >
    > ilyaz.org/software/tmp/KM_A200-resolution-chart-ACRraw-crop-artefacts-10x.png
    >
    > Do you mean (for example) the 4 gray pixels in the horizontal "white"
    > line in the middle of the middle of the image?
    >
    > How I interpret these pixels: there is (effectively) an error in
    > "calculation" of position of straight edges; the error may be as high
    > as 0.25 pixels. I can hardly call this a defect... ;-)
    >
    > (It may be annoying indeed that these "errors" come in periodic
    > groups; still, it is so much sub-pixel...)
    >
    >> I'm viewing on a 19 inch LCD at 1280 x 1024 resolution
    >
    > This is a LCD with 0.29mm dot pitch. On my CRT with 0.26mm dot pitch
    > (actual dot pitch; the monitor's is 0.24mm, IIRC: VS P815) the MTF is
    > so much worse, that I do not see these without magnification...
    >
    > Thanks,
    > Ilya

    Ilya,

    I can't really tell if the pixels you indicated are those I am referring
    to - they may well be. The magnification is so great I can't determine
    which part of the image it is.

    Your comment about visibility is a good one - the MTF of my 19-inch LCD is
    much better than the 19-inch CRT it replaced. Images which looked correct
    before now look quite over sharpened. I can quite accept that on a poorer
    MTF display the eye will average out the pixels so that they are much more
    difficult to display.

    The quarter pixel error in position could lead to a very significant error
    in brightness - I would aim for an error of not more than one tenth of a
    pixel. This is the same sort of error which results from clock timing
    error in audio sampling, and it can sound most unpleasant.

    Cheers,
    David
    Anonymous
    a b C Monitor
    May 31, 2005 1:44:33 AM

    Archived from groups: rec.photo.digital (More info?)

    [A complimentary Cc of this posting was sent to
    David J Taylor
    <david-taylor@blueyonder.co.not-this-bit.nor-this-part.uk>], who wrote in article <K7Cme.42166$G8.15088@text.news.blueyonder.co.uk>:

    > The quarter pixel error in position could lead to a very significant error
    > in brightness - I would aim for an error of not more than one tenth of a
    > pixel. This is the same sort of error which results from clock timing
    > error in audio sampling, and it can sound most unpleasant.

    The difference with audio is that the image which reaches the sensor
    is already severely distorted by diffraction and lens imperfections.
    The Rayleigh's limit for the image is about the pixel pitch; the
    Nyquist limit of the cutoff requency of the diffraction is about 1/3
    of the pixel pitch. Thus I do not think 1/10 pixel errors are not
    practically achievable or even desirable.

    ------

    Now the image: it is of the horizontal lines immediately to the right
    of the center; the left edge is approximately above the (unmarked)
    "19" tick. It is taken from ACRaw link, not from my MTF-compensated
    file.

    Thanks,
    Ilya
    Anonymous
    a b C Monitor
    May 31, 2005 3:11:21 AM

    Archived from groups: rec.photo.digital (More info?)

    Ilya Zakharevich wrote:
    > [A complimentary Cc of this posting was sent to
    > David J Taylor
    > <david-taylor@blueyonder.co.not-this-bit.nor-this-part.uk>], who
    > wrote in article <K7Cme.42166$G8.15088@text.news.blueyonder.co.uk>:
    >
    >> The quarter pixel error in position could lead to a very significant
    >> error in brightness - I would aim for an error of not more than one
    >> tenth of a pixel. This is the same sort of error which results from
    >> clock timing error in audio sampling, and it can sound most
    >> unpleasant.
    >
    > The difference with audio is that the image which reaches the sensor
    > is already severely distorted by diffraction and lens imperfections.
    > The Rayleigh's limit for the image is about the pixel pitch; the
    > Nyquist limit of the cutoff requency of the diffraction is about 1/3
    > of the pixel pitch. Thus I do not think 1/10 pixel errors are not
    > practically achievable or even desirable.
    >
    > ------
    >
    > Now the image: it is of the horizontal lines immediately to the right
    > of the center; the left edge is approximately above the (unmarked)
    > "19" tick. It is taken from ACRaw link, not from my MTF-compensated
    > file.
    >
    > Thanks,
    > Ilya

    The audio comparison is an interesting one, as usually the audio signal is
    properly anti-aliased before digitisation, which is not true in
    photography where the low-pass filter is not as good as in aduio, and may
    even be absent from some cameras. Thus errors in sampling position with
    cameras may be even more signnificant than errors in sampling time with
    audio.

    I still don't know exactly what causes the nasty errors you have produced.
    Could you post an image covering about nine times the area?

    Cheers,
    David
    Anonymous
    a b C Monitor
    May 31, 2005 3:00:23 PM

    Archived from groups: rec.photo.digital (More info?)

    [A complimentary Cc of this posting was sent to
    David J Taylor
    <david-taylor@blueyonder.co.not-this-bit.nor-this-part.uk>], who wrote in article <t4Nme.42526$G8.24244@text.news.blueyonder.co.uk>:

    > I still don't know exactly what causes the nasty errors you have produced.

    I still do not know what you mean by "nasty errors". The stuff I can
    see is very far from being nasty...

    > Could you post an image covering about nine times the area?

    ilyaz.org/software/tmp/KM_A200-resolution-chart-ACRraw-artefact-crop1-10x.png

    Is it the stuff you consider nasty?

    Thanks,
    Ilya
    Anonymous
    a b C Monitor
    May 31, 2005 3:43:19 PM

    Archived from groups: rec.photo.digital (More info?)

    Ilya Zakharevich wrote:
    > [A complimentary Cc of this posting was sent to
    > David J Taylor
    > <david-taylor@blueyonder.co.not-this-bit.nor-this-part.uk>], who
    > wrote in article <t4Nme.42526$G8.24244@text.news.blueyonder.co.uk>:
    >
    >> I still don't know exactly what causes the nasty errors you have
    >> produced.
    >
    > I still do not know what you mean by "nasty errors". The stuff I can
    > see is very far from being nasty...
    >
    >> Could you post an image covering about nine times the area?
    >
    > ilyaz.org/software/tmp/KM_A200-resolution-chart-ACRraw-artefact-crop1-10x.png
    >
    > Is it the stuff you consider nasty?
    >
    > Thanks,
    > Ilya

    Yes, exactly so. I took the region from 9 - 10 (units? I'll call it
    cycles) and resampled it in Paint Shop Pro (no particular algorithm) to
    50% of its size making the spatial frequency cover the 19 cycles value
    where the "nasties" occur. You can see a side-by-side comparison of the
    interpolated 19 cycles and the actual 19 cycles here:

    http://www.david-taylor.myby.co.uk/imaging/sf-9.5.png

    Magnifying to 200% shows the artefacts nicely, and they are clearly
    visible as a roughness in the image on my LCD display.

    By the way: I'm not claiming that Paint Shop Pro does the perfect
    interpolation, simply that it does not have the artefacts of the actual
    image.

    Cheers,
    David
    Anonymous
    a b C Monitor
    June 2, 2005 12:04:09 PM

    Archived from groups: rec.photo.digital (More info?)

    [A complimentary Cc of this posting was sent to
    David J Taylor
    <david-taylor@blueyonder.co.not-this-bit.nor-this-part.uk>], who wrote in article <r5Yme.42673$G8.20710@text.news.blueyonder.co.uk>:
    > Yes, exactly so. I took the region from 9 - 10 (units? I'll call it
    > cycles) and resampled it in Paint Shop Pro (no particular algorithm) to
    > 50% of its size making the spatial frequency cover the 19 cycles value
    > where the "nasties" occur. You can see a side-by-side comparison of the
    > interpolated 19 cycles and the actual 19 cycles here:
    >
    > http://www.david-taylor.myby.co.uk/imaging/sf-9.5.png
    >
    > Magnifying to 200% shows the artefacts nicely, and they are clearly
    > visible as a roughness in the image on my LCD display.
    >
    > By the way: I'm not claiming that Paint Shop Pro does the perfect
    > interpolation, simply that it does not have the artefacts of the actual
    > image.

    What you say and show is essentially:

    the image of a wave with a frequency 76% of the Nyquist looks
    different (and worse!) than one at 38% of the Nyquist.

    Somehow I do not find this any way surprising... What I *do* find
    surprising is that *any* information is preserved at 76% of the
    Nyquist. ;-)

    BTW, about the audio analogy: the cutoff frequency of audio is where
    most (99% or so?) of people would not notice that any cutoff is
    there. The cutoff of current cameras is WAY below this limit.

    E.g., as I wrote in a different thread, the cutoff of 4x5in image with
    Velvia 50 and f/45 (about 50% of 4x5in images are similar to this) is
    close to 80 MPixels. And this is what many would consider as getting
    close to "natural" vision limit...

    Thanks,
    Ilya
    Anonymous
    a b C Monitor
    June 2, 2005 12:36:06 PM

    Archived from groups: rec.photo.digital (More info?)

    Ilya Zakharevich wrote:
    > [A complimentary Cc of this posting was sent to
    > David J Taylor
    > <david-taylor@blueyonder.co.not-this-bit.nor-this-part.uk>], who
    > wrote in article <r5Yme.42673$G8.20710@text.news.blueyonder.co.uk>:
    >> Yes, exactly so. I took the region from 9 - 10 (units? I'll call it
    >> cycles) and resampled it in Paint Shop Pro (no particular algorithm)
    >> to 50% of its size making the spatial frequency cover the 19 cycles
    >> value where the "nasties" occur. You can see a side-by-side
    >> comparison of the interpolated 19 cycles and the actual 19 cycles
    >> here:
    >>
    >> http://www.david-taylor.myby.co.uk/imaging/sf-9.5.png
    >>
    >> Magnifying to 200% shows the artefacts nicely, and they are clearly
    >> visible as a roughness in the image on my LCD display.
    >>
    >> By the way: I'm not claiming that Paint Shop Pro does the perfect
    >> interpolation, simply that it does not have the artefacts of the
    >> actual image.
    >
    > What you say and show is essentially:
    >
    > the image of a wave with a frequency 76% of the Nyquist looks
    > different (and worse!) than one at 38% of the Nyquist.
    >
    > Somehow I do not find this any way surprising... What I *do* find
    > surprising is that *any* information is preserved at 76% of the
    > Nyquist. ;-)
    >
    > BTW, about the audio analogy: the cutoff frequency of audio is where
    > most (99% or so?) of people would not notice that any cutoff is
    > there. The cutoff of current cameras is WAY below this limit.
    >
    > E.g., as I wrote in a different thread, the cutoff of 4x5in image with
    > Velvia 50 and f/45 (about 50% of 4x5in images are similar to this) is
    > close to 80 MPixels. And this is what many would consider as getting
    > close to "natural" vision limit...
    >
    > Thanks,
    > Ilya

    What I show is that the image at 19 cycles from the camera is nothing like
    it should be, as it contains all sorts of nasty artefacts which should not
    be present.

    I agree that on 4x5 the issues may well be different because the sampling
    may be nearer the leans cut-off.

    With audio and sampling position error, the errors result in spurious
    sounds which /are/ within the audible range - i.e. the errors are
    effectively aliased into the audible region. Normally these errors are
    slight, but can result in a veiling of the sound which should be absent on
    better equipment.

    We haven't yet established where these visual artefacts come from, but
    they will result in a roughness on edges in the image, which may lead to
    the image being less pleasing than one taken with better quality (read
    better algorithms?) equipment.

    Cheers,
    David
    Anonymous
    a b C Monitor
    July 5, 2005 10:58:22 PM

    Archived from groups: rec.photo.digital (More info?)

    [Posted with a big delay; my apologies; the thread on
    http://groups-beta.google.com/group/rec.photo.digital/b...
    ]

    I wrote in article <d7mehp$1g9g$1@agate.berkeley.edu>:

    > > Yes, exactly so. I took the region from 9 - 10 (units? I'll call it
    > > cycles) and resampled it in Paint Shop Pro (no particular algorithm) to
    > > 50% of its size making the spatial frequency cover the 19 cycles value
    > > where the "nasties" occur. You can see a side-by-side comparison of the
    > > interpolated 19 cycles and the actual 19 cycles here:
    > >
    > > http://www.david-taylor.myby.co.uk/imaging/sf-9.5.png
    > >
    > > Magnifying to 200% shows the artefacts nicely, and they are clearly
    > > visible as a roughness in the image on my LCD display.
    > >
    > > By the way: I'm not claiming that Paint Shop Pro does the perfect
    > > interpolation, simply that it does not have the artefacts of the actual
    > > image.
    >
    > What you say and show is essentially:
    >
    > the image of a wave with a frequency 76% of the Nyquist looks
    > different (and worse!) than one at 38% of the Nyquist.

    Finally, I think I recognized what is the reason for these artefacts
    (the full picture is on
    http://ilyaz.org/software/tmp/KM_A200-resolution-chart-...
    ).

    DESCRIPTION:

    on high-contrast almost-horizontal and almost-vertical boundaries
    one can see (given extremely good viewing conditions) period=2pixels
    oscillations of brightness when one goes ALONG the edge.

    POSSIBLE REASON: 2 is the period of Bayer sensor.

    Different color cells in Bayer sensor have slightly different point
    spread functions (PSF). Given a sharp boundary black/white, cells
    deep enough in black area are black (no matter what PSF is); likewise
    for pixels deep enough in white area. Moreover, pixels which are
    EXACTLY on the boundary will get middle exposure (for any symmetric
    PSF; and with very high quality lens as on KM A200 it is going to be
    almost symmetric).

    This leaves two narrow (probably less than 1/2-pixel-wide) bands on
    two sides of the black/white bounday where the response of the cell
    depends essentially on the PSF; so slight variations of PSF will
    result in minor variations of the response of the cells. With a
    slightly sloped boundary, sometimes it's offset to the sensor grid is
    going to be close to the "sensitive" offset. In these regions some
    artefacts similar to the discussed above are going to be observable.

    HOW TO FIX:

    The demosaicer should treat different positions in Bayers sensor
    differently. The Adobe demosaicer already deconvolves to compensate
    for an "average" PSF (as the calculated throughput MTF shows); it
    should do more.

    WHY DIFFERENT PSF:

    Two reasons for variation of PSF for different types of cells:

    a) chromatic aberration (while almost unnoticable on KM A200 in
    other situations, it may still contribute to this effect);

    b) Photons of different wavelength are converted to electrons/holes
    pairs at different depth in silicon; then electrons/holes travel
    across silicon to discharge the capacitors of cells. The pairs
    generated at larger depth need to travel more, so they have a
    larger chance to diffuse to the neighbour cell meanwhile.

    CONCLUSION:

    If my estimates are correct, this particular type of artefacts should
    be fixable by an update to demosaicer.

    Hope this helps,
    Ilya
    Anonymous
    a b C Monitor
    July 6, 2005 1:12:23 PM

    Archived from groups: rec.photo.digital (More info?)

    Ilya Zakharevich wrote:
    []
    > HOW TO FIX:
    >
    > The demosaicer should treat different positions in Bayers sensor
    > differently. The Adobe demosaicer already deconvolves to compensate
    > for an "average" PSF (as the calculated throughput MTF shows); it
    > should do more.
    >
    > CONCLUSION:
    >
    > If my estimates are correct, this particular type of artefacts should
    > be fixable by an update to demosaicer.
    >
    > Hope this helps,
    > Ilya

    You may well be correct. My original encounter with these artefacts was
    with the Minolta A1 camera, when DP Review reported them. I actually
    bought a Minolta A2 camera, blindly hoping that the problem would have
    been fixed in the newer firmware, but it had not. The camera went back,
    of course.

    The issue was that JPEGs direct from the camera had these artefacts,
    whereas as RAW files converted on the computer did not. This does indeed
    suggest that the in-camera demosaic algortihm was not as good as the
    external software, and that the algorithm differences produced the
    artefacts. To me, it still looks like integer overflow and underflow,
    rather than being simply a small variation caused by PSF differences, but
    the two may well be related.

    The essence appears to be (at least in the Minolta A1 and A2) that the
    firmware in the camera doesn't do a good enough job.

    Thanks for continuing to check out this problem.

    Cheers,
    David
    Anonymous
    a b C Monitor
    July 7, 2005 8:13:20 AM

    Archived from groups: rec.photo.digital (More info?)

    [A complimentary Cc of this posting was sent to
    David J Taylor
    <david-taylor@blueyonder.co.not-this-bit.nor-this-part.uk.invalid>], who wrote in article <XfNye.63897$G8.13982@text.news.blueyonder.co.uk>:

    > > If my estimates are correct, this particular type of artefacts should
    > > be fixable by an update to demosaicer.

    > You may well be correct. My original encounter with these artefacts was
    > with the Minolta A1 camera, when DP Review reported them. I actually
    > bought a Minolta A2 camera, blindly hoping that the problem would have
    > been fixed in the newer firmware, but it had not. The camera went back,
    > of course.
    >
    > The issue was that JPEGs direct from the camera had these artefacts,
    > whereas as RAW files converted on the computer did not. This does indeed
    > suggest that the in-camera demosaic algortihm was not as good as the
    > external software, and that the algorithm differences produced the
    > artefacts.

    Just for the record, let me repeat that I could not agree with your
    points less...

    a) I do not consider this artefact as a defect in A200;

    a') All the digital cameras I have seen (those with high-quality lens)
    show exactly the same artefact; the degree by which it is visible
    is determined mostly by how aggressively the camera/demosaicer
    kill information in high-frequency region.

    b) This issue is NOT of in-camera vs. standalone demosaicer. I did
    not hear of ANY demosaicer in existence which compensates for this
    artefact.

    I do not think that any firmware on a camera in the next couple of
    years is going to have enough processing power to do compensation
    of this artefact.

    [Hmm, on the other hand it is not a very complicated convolution...
    But still, since Adobe does not do it, I feel safe predicting that
    no in-camera firmware is going to do it soon.]

    c) This artefact is MORE visible on RAW images than on JPEG; the
    reason is that RAW images usually have much better high-frequency
    contrast.

    Hope this helps (and hope that we are discussing the same artefacts ;-),
    Ilya
    Anonymous
    a b C Monitor
    July 7, 2005 12:22:00 PM

    Archived from groups: rec.photo.digital (More info?)

    Ilya Zakharevich wrote:
    > [A complimentary Cc of this posting was sent to
    > David J Taylor
    > <david-taylor@blueyonder.co.not-this-bit.nor-this-part.uk.invalid>],
    > who wrote in article
    > <XfNye.63897$G8.13982@text.news.blueyonder.co.uk>:
    >
    >>> If my estimates are correct, this particular type of artefacts
    >>> should be fixable by an update to demosaicer.
    >
    >> You may well be correct. My original encounter with these artefacts
    >> was with the Minolta A1 camera, when DP Review reported them. I
    >> actually bought a Minolta A2 camera, blindly hoping that the problem
    >> would have been fixed in the newer firmware, but it had not. The
    >> camera went back, of course.
    >>
    >> The issue was that JPEGs direct from the camera had these artefacts,
    >> whereas as RAW files converted on the computer did not. This does
    >> indeed suggest that the in-camera demosaic algortihm was not as good
    >> as the external software, and that the algorithm differences
    >> produced the artefacts.
    >
    > Just for the record, let me repeat that I could not agree with your
    > points less...

    OK, then will have to agree to differ.

    > a) I do not consider this artefact as a defect in A200;

    Any artefact is a defect, certainly when it is produced by incorrect
    processing.

    > a') All the digital cameras I have seen (those with high-quality lens)
    > show exactly the same artefact; the degree by which it is visible
    > is determined mostly by how aggressively the camera/demosaicer
    > kill information in high-frequency region.

    I have seen cameras with a similar basi specification produce differing
    degrees of such artefacts. Those with less artefact I consider better
    designed.

    > b) This issue is NOT of in-camera vs. standalone demosaicer. I did
    > not hear of ANY demosaicer in existence which compensates for this
    > artefact.

    It certainly is on the Minolta A1 and A2, where the in-camera software
    produces the problem, but the computer software for converting RAW files
    does not. This was commented on in the DPReview article.

    > I do not think that any firmware on a camera in the next couple of
    > years is going to have enough processing power to do compensation
    > of this artefact.
    >
    > [Hmm, on the other hand it is not a very complicated convolution...
    > But still, since Adobe does not do it, I feel safe predicting that
    > no in-camera firmware is going to do it soon.]

    No comment.

    > c) This artefact is MORE visible on RAW images than on JPEG; the
    > reason is that RAW images usually have much better high-frequency
    > contrast.

    I can't comment on that.

    > Hope this helps (and hope that we are discussing the same artefacts
    > ;-),
    > Ilya

    Cheers,
    David
    !