We need standards
Mark Herring
I would like to open a dialog on how we can move towards some
standards for how digital imaging technology is specified. The
regulars here will quickly see that I am motivated in large part by
the rather amazing debates about the merits of the FoveonX3 and Fuji
"Super-CCD" technology.
The most urgent need, in my view, is to somehow capture what I have
called the "de-facto" standard for how spatial resolution is
specified---ie the number of true, spatially independent, samples that
are acquired by the sensor. It is an imperfect definition, in that it
only gives you the limiting resolution---ie the best that the device
can be if all other factors are perfect. It does, however, allow some
rational basis for comparing cameras.
In addition to standards for specifications, would it not also be
useful to have some standard tests---ones that could be duplicated
anywhere? Such a test would require that something be photographed
under controlled conditions. The "penny test" is a step in this
direction, but random pennies pulled out of one's pocket are all
different, and the controlled conditions are not specified.
Does anyone know if any of the traditional organizations---eg ISO--are
dealing with this issue?...or plan to?
Any ideas for what the standards should be?
-Mark
**************************
Mark Herring, Pasadena, Calif.
Private e-mail: Just say no to "No".
Tony Spadaro
to inflate sales will ignore those stnadards and continue to come to the
conclusion that 3.4 equals 10.
Thirty years ago there were a half dozen ways to measure the output of a
HiFi amplifier. After a long battle the RMS system was declared to tbe the
"standard" by which all such amplifiers are measuered. Today the boxes
simply state an output in watts - with no explanation of what measurement
was used. Companies simply ignored the RMS "standard" and put up any
inflated number they pleased.
--
http://www.chapelhillnoir.com
home of The Camera-ist's Manifesto
The Improved Links Pages are at
http://www.chapelhillnoir.com/links/mlinks00.html
A sample chapter from my novel "Haight-Ashbury" is at
http://www.chapelhillnoir.com/writ/hait/hatitl.html
"Mark Herring" <Nom...@surfcity.net> wrote in message
news:s04eqv0009raucj91...@4ax.com...
Samuel Paik
> The most urgent need, in my view, is to somehow capture what I have
> called the "de-facto" standard for how spatial resolution is
> specified---ie the number of true, spatially independent, samples that
> are acquired by the sensor. It is an imperfect definition, in that it
> only gives you the limiting resolution---ie the best that the device
> can be if all other factors are perfect. It does, however, allow some
> rational basis for comparing cameras.
Standard? It's pretty easy to get these numbers for most digital
image sensors. The oddball one is for the various Fuji "weirdos".
> In addition to standards for specifications, would it not also be
> useful to have some standard tests---ones that could be duplicated
> anywhere?
There are standard ISO test targets. Both Dpreview and Imaging
Resource routinely take photos of the basic resolution chart
in their reviews, although they don't set up completely precisely,
they do a reasonable enough job to give readers a good idea.
(BTW, the basic chart is inadequate--does not supply close enough
lines--for the highest resolution DSLRs available today, e.g.
EOS-1Ds, Kodak DCS-14n).
Various organizations have even more sophisticated tests available.
Don Stauffer
one uses it. Mfgrs sure don't. And mags reporting on cameras do not
do the test specified in the standard. I think the problem is, how do we
get people to use the existing standard.
--
Don Stauffer in Minnesota
stau...@usfamily.net
webpage- http://www.usfamily.net/web/stauffer
Bart van der Wolf
> Does anyone know if any of the traditional organizations---eg ISO--are
> dealing with this issue?...or plan to?
In addition to available standards, the ISO is finalizing several standards
for "Electronic still-picture cameras". But of course, anyone can choose to
ignore them.
> Any ideas for what the standards should be?
ISO 12233:2000 Photography -- Electronic still-picture cameras --
Resolution measurements (available in English only):
(http://www.iso.org/iso/en/CatalogueDetailPage.CatalogueDetail?CSNUMBER=3371
5&ICS1=37&ICS2=40&ICS3=10).
For CHF 114,00 it's yours (PDF or paper version).
The ISO is very clear about pixels not being a measure of resolution, which
makes sense because a pixel is dimensionless until output to a physical
medium. When pixels are mentioned, they are often referred to as "luminance
pixels".
They do specify the following:
- Resolution: a measure of the ability of a digital image capture system,
or a component of a digital image capture system, to capture fine spatial
detail. Resolution measurement metrics include resolving power, limiting
visual resolution, SFR, MTF and CTF.
- Sampled imaging system: an imaging system or device which generates an
image signal by sampling an image at an array of discrete points, or along a
set of discrete lines, rather than a continuum of points. The sampling at
each point is done using a finite size sampling aperture or area.
- Sample spacing: the physical distance between sampling points or sampling
lines. The sample spacing may be different in the two orthogonal sampling
directions. Measured in units of distance (e.g., microns, mm.)
- Sampling frequency: the reciprocal of sample spacing. Expressed in samples
per unit distance. (e.g. dots per inch, DPI)
Standards also relevant for this subject are:
ISO 12231:1997 Photography -- Electronic still-picture cameras --
Terminology (available in English only). Note: A new version is in the works
(ISO/DIS 12231-2)
ISO 14524:1999 Photography -- Electronic still-picture cameras -- Methods
for measuring opto-electronic conversion functions (OECFs) (available in
English only)
ISO 15739:2003 Photography -- Electronic still-picture imaging -- Noise
measurements (available in English only)
and for current projects:
<http://www.iso.ch/iso/en/stdsdevelopment/techprog/workprog/TechnicalProgram
meTCDetailPage.TechnicalProgrammeTCDetail?COMMID=1603>
Careful, the link probably wraps to two lines.
Bart
Mark Herring
<stau...@usfamily.net> wrote:
>There is a fine standard for just that, an ISO standard. Problem is, no
>one uses it. Mfgrs sure don't. And mags reporting on cameras do not
>do the test specified in the standard. I think the problem is, how do we
>get people to use the existing standard.
>
Don;
What standard are you referring to?
What I am looking for is something simple like ISO film speeds. ISO
400 gives my a pretty good picture of what a film will do.
Just as an example, If we had a standard for maximum resolution that
was unambiguous, then it might not be that hard to enforce it--at
least in reviews, newgroup discussions, etc.
_Mark
Mark Herring
<bvd...@no.spam> wrote:
>ISO 12233:2000 Photography -- Electronic still-picture cameras --
>Resolution measurements (available in English only):
>(http://www.iso.org/iso/en/CatalogueDetailPage.CatalogueDetail?CSNUMBER=3371
>5&ICS1=37&ICS2=40&ICS3=10).
>For CHF 114,00 it's yours (PDF or paper version).
>
Lots snipped.......
>
>and for current projects:
><http://www.iso.ch/iso/en/stdsdevelopment/techprog/workprog/TechnicalProgram
>meTCDetailPage.TechnicalProgrammeTCDetail?COMMID=1603>
>Careful, the link probably wraps to two lines.
>
>Bart
>
Bart;
Thanks for all the info---can't absorb it at the moment. In all this,
do you see any simple standard for resolution---akin to ISO film
speed?
Don Stauffer
quantified ability to do a specific task task, and is the result of
tests. Normally in photographic and TV work, it is the ability to
recognize that a series of alternating dark and light bars are just
that, not a grey patch.
Bart van der Wolf wrote:
>
> The ISO is very clear about pixels not being a measure of resolution, which
> makes sense because a pixel is dimensionless until output to a physical
> medium. When pixels are mentioned, they are often referred to as "luminance
> pixels".
>
>
> Bart
Frank ess
> On Tue, 4 Nov 2003 15:38:08 +0100, "Bart van der Wolf"
> <bvd...@no.spam> wrote:
>
>
> >ISO 12233:2000 Photography -- Electronic still-picture cameras --
> >Resolution measurements (available in English only):
>
>(http://www.iso.org/iso/en/CatalogueDetailPage.CatalogueDetail?CSNUMBER=337
1
> >5&ICS1=37&ICS2=40&ICS3=10).
> >For CHF 114,00 it's yours (PDF or paper version).
> >
>
> Lots snipped.......
> >
> >and for current projects:
>
><http://www.iso.ch/iso/en/stdsdevelopment/techprog/workprog/TechnicalProgra
m
> >meTCDetailPage.TechnicalProgrammeTCDetail?COMMID=1603>
> >Careful, the link probably wraps to two lines.
> >
> >Bart
> >
> Bart;
>
> Thanks for all the info---can't absorb it at the moment. In all this,
> do you see any simple standard for resolution---akin to ISO film
> speed?
>
I think it is on the schedule just after "audio-Watts-of-power" is resolved.
TheWeb...@webtv.net
choose.
Bart van der Wolf
> On Tue, 4 Nov 2003 15:38:08 +0100, "Bart van der Wolf"
> <bvd...@no.spam> wrote:
>
>
> >ISO 12233:2000 Photography -- Electronic still-picture cameras --
> >Resolution measurements (available in English only):
> Thanks for all the info---can't absorb it at the moment. In all this,
> do you see any simple standard for resolution---akin to ISO film
> speed?
The ISO "slanted edge" method is pretty simple, it only needs some
mathematical evaluation to get the results (MTF like graphs, called SFR or
Spatial Frequency Response curves). It will take some programming to speed
up the evaluation. I may be able to find a method using free software.
An alternative can be home made with little effort. I use a relatively
simple star target, that gives both a visual impression, and the possibility
to calculate the limiting resolution or an MTF (requires to take multiple
measurements). It is easy to set up, use, and the interpretation is also
very simple.
Bart
Bart van der Wolf
"Bart van der Wolf" <bvd...@no.spam> wrote in message
news:3fa98ed7$0$58702$e4fe...@news.xs4all.nl...
SNIP
> An alternative can be home made with little effort. I use a relatively
> simple star target, that gives both a visual impression, and the
possibility
> to calculate the limiting resolution or an MTF (requires to take multiple
> measurements). It is easy to set up, use, and the interpretation is also
> very simple.
I've just uploaded the file (2 versions) for download.
For HP inkjet printers (3.77MB):
http://www.xs4all.nl/~bvdwolf/main/downloads/Jtf60cy-100mm_600ppi.gif
For Epson inkjet printers (5.28MB):
http://www.xs4all.nl/~bvdwolf/main/downloads/Jtf60cy-100mm_720ppi.gif
Print this file unaltered (switch-off printer enhancements) at the indicated
ppi setting (use glossy paper for output). It should result in a print of
100x100mm.
Shoot an image of the target from a couple of meters distance (exact
distance is not critical, 2-3 meters is fine). Using a tripod, mirror
lock-up, and stopping down your lens to the optimal aperture will benefit
the result. You may want to shoot several frames to eliminate autofocus
variance and afterwards choose the best one (digital is cheap).
Evaluation:
1. First you need to determine the sampling pitch (center to center distance
between photosites in a row or column) of your sensor.
For example, if you scan film with a 4000ppi scanner, the pitch is 1/4000
inch, or 25.4/4000=0.00635 mm per output pixel, if you use a Canon D60 /
10D, then the pitch is about 0.007377 mm, a 1Ds has a pitch of 0.0088 mm, a
Sigma SD9/SD10 has a pitch of 0.00912 mm, etc., etc..
2. Open the image in your photoeditor, and zoom in (e.g. 300%) on the
target. You can approximately color balance the target by setting the
blackest point in your target image close to RGB 5,5,5 and the whitest point
to RGB 240,240,240. Adjusting the midpoint (target background) to RGB
127,127,127 should then give a neutral balance with approximately the right
contrast for the entire image.
3. The star target image consists of 60 full modulation cycles from black to
white, on a mid gray background. The circumference of any concentric circle
will also equal 60 cycles. Taking Pi (3.14159...) and the pitch in
millimeters you determined in step 1, this means that for your sensor 1
cycle/mm = 60/(Pi*pitch) pixels (you may approximate that with 19.1/pitch).
In your photo editor you now measure the diameter of the unresolved center
in pixels. Now divide the number of pixels you found for 1 cy/mm by the
diameter in pixels you measured, and the result is the limiting resolution
in cy/mm or lp/mm for your entire imaging chain.
As long as you shoot with the same sensor pitch, you can use the same
calculation and only divide by the newly measured diameter. The blur
diameter on the sensor will not change with target distance, because the
resolution is the limiting factor and not the target magnification.
This will enable to find the differences between lenses, zoom settings, and
apertures for *your* imaging chain. It will also help to estimate the amount
of magnification you can apply to your images before the viewing resolution
is compromised.
Bart
David J. Littleboy
"Bart van der Wolf" <bvd...@no.spam> wrote:
>
> I've just uploaded the file (2 versions) for download.
> For HP inkjet printers (3.77MB):
> http://www.xs4all.nl/~bvdwolf/main/downloads/Jtf60cy-100mm_600ppi.gif
> For Epson inkjet printers (5.28MB):
> http://www.xs4all.nl/~bvdwolf/main/downloads/Jtf60cy-100mm_720ppi.gif
Thanks!
--
David J. Littleboy
dav...@gol.com
Tokyo, Japan