Lens profiling tool from Adobe ( for CS5 / LR / ACR ).

Annika1980 wrote:
On May 11, 11:00 pm, "Bill Graham" wrote:

The anti distortion software could be built
right into the camera. In time, virtually all aberrations peculiar to lenses
could be solved in software, and built into the cameras, couldn't they?

Canon has already started doing that with it's Peripheral Illumination
Correction built into the Fab 5D2 and the 50D. I suppose in future
releases we will see correction for other lens aberrations.

http://www.usa.canon.com/dlc/control...articleID=2027

So has Nikon but I wish they'd make it more accessible to raw converters
because that's worthless shooting raw without their (paid) software.
grumble

"Mxsmanic" wrote in message
...
David J Taylor writes:

IASI:

http://smsc.cnes.fr/documentation/IA.../SPIE_ASPI.pdf

Still looks like one photosite = one number. Three detectors provide
three
(optically coincident?) photosites, each of which provides one signal
level
per pixel, if I'm reading the poorly-written publication correctly.

The spectrum at each pixel is analysed, producing over 8000 value per
pixel. I looked for a better description but didn't find one quickly.

Cheers,
David

"David J Taylor" wrote:
"Mxsmanic" wrote in message
.. .
David J Taylor writes:

IASI:

http://smsc.cnes.fr/documentation/IA.../SPIE_ASPI.pdf

Still looks like one photosite = one number. Three
detectors provide three
(optically coincident?) photosites, each of which
provides one signal level
per pixel, if I'm reading the poorly-written publication correctly.

The spectrum at each pixel is analysed, producing over 8000 value per
pixel. I looked for a better description but didn't find one quickly.

I read through the cite provided, and saw nothing
indicating 8000 values per pixel. It might well be that
the final product distributed to others does indeed have
that resolution, just that the cited paper did not say
that.

What it does say is that each set of pixel data from the
sensor is a 3 channel 16 bit sample. That is, 48 bits,
or 2.8e+14 values per bit.

Regardless of that it is a terrific example to show what
is wrong with the various naive statements made by
Mxsmaniac on this topic. If one wants to build a
sensor/camera to produce photographs that will be
similar to what the human eye can see, the technology
and engineering will necessarily be distinctly different
than a sensor/camera that is designed to produce
spectrographics. They both do basically the same
things, and use similar modular parts (encoding filters,
sensors, ADC's, etc), but the parameters for each are
necessarily very different to match the design target.

Interferometers rather than Bayer filters, slower ADC's
with multiple scans at different ISO gains to achieve
higher bit density, and an entirely different post
processing toolset are the obvious differences.

But they are all equivalents, the data chain is the
same, and it is clear that three channel "color" sensor
data can indeed define a continuous spectrum of
wavelengths, even into the IR region.


--
Floyd L. Davidson http://www.apaflo.com/floyd_davidson
Ukpeagvik (Barrow, Alaska)

On 5/15/2010 2:23 AM, David J Taylor wrote:
"Mxsmanic" wrote in message
...
David J Taylor writes:

IASI:

http://smsc.cnes.fr/documentation/IA.../SPIE_ASPI.pdf

Still looks like one photosite = one number. Three detectors provide
three
(optically coincident?) photosites, each of which provides one signal
level
per pixel, if I'm reading the poorly-written publication correctly.

The spectrum at each pixel is analysed, producing over 8000 value per
pixel. I looked for a better description but didn't find one quickly.

The authors are writing for a specialist audience, people who do optical
interferometry or design satellites for a living. They're assuming that
one already understands the technology and are pointing out the details
of their particular implmentation.

Another paper, describing an earlier satellite that uses a similar
sensor, is http://www.atmos-chem-phys.net/8/2151/2008/acp-8-2151-2008.pdf.

The keyword you need is "FTIR" or "Fourier Transform Infrared
spectrometer". There's a tutorial at
http://www2.fc.up.pt/pessoas/peter.eaton/tutorial/webCT/index.html
that includes a brief discussion of the technology and instructions for
operating a specific instrument. A more detailed discussion can be
found at http://spectroscopy.lbl.gov/FTIR-Martin/

Saying that "the spectrum is analyzed" implies that some kind of magic
is done with a single data point to derive 8000 values. It doesn't work
that way--the whole system is purpose made to provide a spectrum, using
moving parts and series of measurements, it's not something done in post
as an afterthought. An FTIR uses a moving mirror to generate a range of
values over time by altering the optical path length. What you get is
the fourier transform of the spectrum--you have to do calculations on
that to get the actual spectrum. The whole thing is really quite clever.

At each point in time the sensor gives a single value. Over a movement
cycle of the mirror it will give some wide range of them, what that
range is, specifically, will depend on the rate at which the mirror is
moving and the response time and resolution of the sensor. But all the
sensor measures is intensity.

"Paul Furman" wrote in message
...
Annika1980 wrote:
On May 11, 11:00 pm, "Bill Graham" wrote:

The anti distortion software could be built
right into the camera. In time, virtually all aberrations peculiar to
lenses
could be solved in software, and built into the cameras, couldn't they?

Canon has already started doing that with it's Peripheral Illumination
Correction built into the Fab 5D2 and the 50D. I suppose in future
releases we will see correction for other lens aberrations.

http://www.usa.canon.com/dlc/control...articleID=2027

So has Nikon but I wish they'd make it more accessible to raw converters
because that's worthless shooting raw without their (paid) software.
grumble


I share your wish. But, I don't agree though that raw shooting makes no
sense unless you use CaptureNX. Raw still allows a lot more adjustments in
PS, with significantly less data loss than jpeg or tiff.


--
Peter

On Fri, 14 May 2010 22:43:14 -0700, Paul Furman
wrote:

So has Nikon but I wish they'd make it more accessible to raw converters
because that's worthless shooting raw without their (paid) software.
grumble


Bibble has had lens correction for a while. First PTLens, then it's
own implementation.

In message , Paul Furman paul-
@-edgehill.net writes
Annika1980 wrote:
On May 11, 11:00 pm, "Bill Graham" wrote:

The anti distortion software could be built
right into the camera. In time, virtually all aberrations peculiar to lenses
could be solved in software, and built into the cameras, couldn't they?
Canon has already started doing that with it's Peripheral
Illumination
Correction built into the Fab 5D2 and the 50D. I suppose in future
releases we will see correction for other lens aberrations.
http://www.usa.canon.com/dlc/control...Act&articleID=
2027

So has Nikon but I wish they'd make it more accessible to raw
converters because that's worthless shooting raw without their (paid)
software. grumble

There is a whole load of software that works with Nikon Raw. What are
you grumbling about?

--
\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\
\/\/\/\/\ Chris Hills Staffs England /\/\/\/\/
\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/

Mxsmanic wrote:
Floyd L. Davidson writes:

What it does say is that each set of pixel data from the
sensor is a 3 channel 16 bit sample. That is, 48 bits,
or 2.8e+14 values per bit.

So three numbers, just like RGB. That's hardly going to record a full
spectrum.

Wrong.

Regardless of that it is a terrific example to show what
is wrong with the various naive statements made by
Mxsmaniac on this topic. If one wants to build a
sensor/camera to produce photographs that will be
similar to what the human eye can see, the technology
and engineering will necessarily be distinctly different
than a sensor/camera that is designed to produce
spectrographics. They both do basically the same
things, and use similar modular parts (encoding filters,
sensors, ADC's, etc), but the parameters for each are
necessarily very different to match the design target.

A spectrograph needs to scan the entire spectrum for each pixel, producing a
very large set of individual numbers that represent the signal strength at
each of many frequencies. It cannot capture the original spectrum completely
without an infinite number of samples, but it can do vastly better than a
single photosite on a digital camera imaging sensor.

In that particular case they were doing 55,000 samples.
You either didn't read the cited article or couldn't
understand it.

The photosite of a sensor produces only one number, which represents the
integration of all frequencies present in the original scene, the sum of the
products of signal strength in the original scene at each frequency and the
sensitivity at that frequency in the sensor. This integration produces just
one number, and it's a one-way transformation. It is impossible to reconstruct
the original spectral distribution of the scene from this single number,
because multiple distributions can produce the same number.

Wrong.

In fact, the spectrograph is the inverse of the digital imaging sensor. The
spectrograph produces a very large set of numbers for the entire image as a
whole, giving its overall frequency distrubtion--but it provides no detail
information, effectively treating the entire image as a single pixel. The
sensor in a camera is exactly the opposite, capturing a great many pixels in
order to record detail, but recording only a tiny fraction of the spectral
data for each pixel.

Hence it is abjectly ridiculous of you to claim that
because a photographic camera does not produce
spectrographs that it cannot see and respond to the
entire spectrum of light.

With a spectrograph, you can reverse the process and get a pretty good
approximation of the original spectral distribution, but you get no detail,
just a complete blur (just one pixel). With a digital camera, you can reverse
the process and get a pretty good approximation of the original image detail,
but virtual all the spectral information is gone.

Wrong.

The digital camera works because human beings concentrate on detail, not
spectra. But the fact remains that, once the image is captured, virtually all
the spectral information is gone, which means that anything that depends on
the original spectrum cannot be simulated in post production.

Again, wrong. Once the the *data* is capture, it is used to produce
a photographic image. It could just as well be used to produce a
spectrograph.

This also applies to certain lens effects, such as blurring and transmission
characteristics, or anything that removes information from the original scene.

Well, isn't that sort of a broad statement if we've ever seen one!
And totally meaningless, too!

But they are all equivalents, the data chain is the
same, and it is clear that three channel "color" sensor
data can indeed define a continuous spectrum of
wavelengths, even into the IR region.

Nope. With only three numbers, a many-into-one conversion of all possible
spectra into all possible number sets occurs, and it is irreversible, so the
original spectrum cannot be recreated.

Untrue.

--
Floyd L. Davidson http://www.apaflo.com/floyd_davidson
Ukpeagvik (Barrow, Alaska)

Bruce wrote:
On Thu, 13 May 2010 22:47:52 -0700, "Bill Graham"
wrote:
I was not aware that there had been an increase in the sales of tilt-shift
lenses in recent years, but there could be several reasons for
this.....Perhaps the price has come down a lot? Or perhaps there has been a
resurgence in architectural photography. Or perhaps there are other things
that tilt-shift lenses will do that I am unaware of that can't be done in
Photoshop


Yes, that's it. Correctly used, a tilt-shift lens can provide
exceptional depth of field that would be virtually impossible to
replicate in software. You can achieve quite a lot by shooting
several shots of the same scene focused at different distances, and
then combining elements of those images, but the results look very
different and quite artificial.

Focus stacking works fine at non-macro distances. It's rare to see
people use that method but it works great.


As for using Photoshop to eliminate converging verticals, as Mxmanic
has explained (correctly, but too many times!) there is a huge loss of
image quality. I am very well acquainted with the picture editor of
an architectural magazine He summarily rejects any shot submitted for
publication where converging verticals have been corrected in
software. He says he can spot them a mile away, and I believe him. I
have tried very hard to hoodwink him but without success ...

I suppose there are subtle cues, but a small amount of correction is
hardly noticeable. The bigger problem is without a shift lens, you can't
frame the composition and don't know what it's going to look like till
later. Shift lenses also have a trade-off of resolution because you are
using the edge of a larger image circle, it will be a little softer and
show some vignetting.

--
Paul Furman
www.edgehill.net
www.baynatives.com

all google groups messages filtered due to spam

"Mxsmanic" wrote in message
...
David J Taylor writes:

The spectrum at each pixel is analysed, producing over 8000 value per
pixel.

Eight thousand separate numbers, or a number that can have 8000
different
values? In the first case, explain how it is doing this with only three
sensors.

Please take a look at J. Clarke's explanation of "Fourier Transform
Infrared
spectrometer". It's not a three sensor camera.

Cheers,
David