Velocity Reviews > Photography Of The Future (Terapixel Technology!)

# Photography Of The Future (Terapixel Technology!)

Michael A. Covington
Guest
Posts: n/a

 11-17-2004
"Matt" <(E-Mail Removed)> wrote in message
news:cne22e\$rtf\$(E-Mail Removed)...
>A guy called Michael, recently replied to a post of mine and said 'How many
> megapixels do we need?'. I thought this was a very good point, and made
> me
> think a bit. (Bear with me on this.)

That's me... I'm a college professor... they pay me to make people think

In what follows, both the wave and the quantum properties of light are going

> Well, picture this. You are a sports photographer and are sitting
> at the boundaries. You have only one lens, which is a wide angle lens.
> You ask, 'Why no telephoto lens? The answer is, 'Because you don't need
> it!'
>
> The reason why you don't need it is because you just shoot away covering
> the
> whole scene you want to shoot, and crop the part of the photo you want
> because you have a gigapixel, or terapixel body.

A body with a 1-million-by-1-million array of pixels, you mean, or something
like that?

Let's do some rough back-of-the-envelope calculations.

Suppose the sensor is 2 inches (50 mm) square. (I'll admit you probably
want the camera to be bigger than 35 mm, but not gigantically bigger.)

Then each pixel is 50 nanometers square.

Guess what? That's 1/10 the wavelength of light! You won't get an image
anywhere *near* that sharp. Every photon will spread over some dozens of
pixels.

And then there's diffraction in the lens. Suppose you have an f/1 lens of
50 mm focal length. (You said wide angle.) Then its aperture is also 50 mm
and its Dawes limit (close to the Rayleigh limit) is 2.25 arc-seconds or
1/100000 radian (roughly). That translates to 500 nanometers.

So again, the resolution of your f/1 lens (even if someone invents a new f/1
lens design that is limited ONLY by diffraction, magically making all
aberrations go away) will be only 1/10 as good as your pixel array.

In practice, lenses are never diffraction limited at f/1. Designing a lens
is a matter of solving simultaneous equations and there hasn't been
tremendous progress in this recently. (There are real mathematical limits
on what can be done.) So allow another factor of 10 for the resolution to
be worse than that.

So now our 1,000,000-by-1,000,000-pixel array is now down to something like
10,000 by 10,000. That's a whopping 100 megapixels, just a factor of 10
more than we commonly have today.

In short -- Your dream sensor is going to be about like Technical Pan Film,
and your lens is going to be about like the best products presently
available from Nikon, Canon, or Zeiss.

It's fun to dream that all technology will advance the way silicon chips
have during the past 40 years. But that's just dreaming if you don't take
into account the underlying physics.

Michael Covington
Associate Director, Artificial Intelligence Center
The University of Georgia - www.ai.uga.edu/mc

Michael A. Covington
Guest
Posts: n/a

 11-17-2004

"Matt" <(E-Mail Removed)> wrote in message
news:cne553\$lvg\$(E-Mail Removed)...
> Well, Karl, don't get me started on medium/large format of the future.
>
> The digital equivalent will be on satellites. This means that it will be
> possible to take photos of the world, and crop to the part you want!
>
> Just a thought, but is there any reason why this technology doesn't
> exist, and is already being used by the government?

Physics. There are real limits to the resolution you can get with a camera
of reasonable size (even ten feet in diameter) due to diffraction, and
there's also the unsteadiness of the air.

Why do we get better pictures of Jupiter with the Hubble Space Telescope
than with ground-based telescopes? Because we don't have to take pictures
through the air. A satellite trying to take a picture of the ground has the
same problem as a telescope trying to look up and out through the air. The
air is very unsteady. Resolution better than 0.5 arc-second is rarely
achievable regardless of telescope size, and much of the time, it's ten
times worse than that.

0.5 arc-second is roughly 1/400000 radian, or the resolution needed to read
a newspaper (distinguishing 0.5mm features) from 200 yards away.

Michael A. Covington
Guest
Posts: n/a

 11-17-2004

"Doc" <(E-Mail Removed)> wrote in message
news:_%wmd.1312\$(E-Mail Removed) nk.net...
>
> "Matt" <(E-Mail Removed)> wrote in message
> news:cne553\$lvg\$(E-Mail Removed)...
>> Well, Karl, don't get me started on medium/large format of the future.
>>
>> The digital equivalent will be on satellites. This means that it will be
>> possible to take photos of the world, and crop to the part you want!
>>
>> Just a thought, but is there any reason why this technology doesn't

>> exist, and is already being used by the government?

>
> Isn't there an optic limit to what you can achieve this way? Just the same
> way that you can't simply use ever more powerful multiplying lenses to
> boost
> the magnification of a given size telescsope (though this is the
> fraudulent
> basis of most consumer telescope power claims). The image loses clarity
> once
> you get past a mathematically determined limit.

Bingo. Diffraction limits. Exactly.

Michael A. Covington
Guest
Posts: n/a

 11-17-2004
"Alan Meyer" <(E-Mail Removed)> wrote in message
news:qaxmd.1962\$(E-Mail Removed)...

> Of course any lens will resolve an image for an infinite
> number of pixels. All you need is an infinite amount of
> exposure time

Not strictly true, but it does touch on another point. If you have tiny,
tiny pixels on the sensor, then it's going to take longer to get a
reasonable number of photons into each of them. Astronomers are already
familiar with this.

BillyJoeJimBob
Guest
Posts: n/a

 11-17-2004
Matt wrote:
> A guy called Michael, recently replied to a post of mine and said 'How many
> megapixels do we need?'. I thought this was a very good point, and made me
> think a bit. (Bear with me on this.)

Oh dear.

> Resolution is something which will constantly increase, just like RAM and
> Processor Speed of computers. If I remember correctly, Bill Gates once said
> something along the lines of '512k RAM is enough for anybody'.

"640k ought to be enough for anybody."

Resolution is not something which will constantly increase. Even with
an infinite pixel sensor you then become limited by the optics in front.

> So, when is enough, enough when it comes to megapixels?

What's the application?

For my purposes it's when the size of the individual pixels, when
printed or displayed and then viewed at the appropriate/desired distance
exceeds the resolving capability of the viewer's eyes. Anything beyond
that is essentially wasted information except for the possibility of
noise reduction which could have been done ahead of time by using
fewer, larger, pixels in the sensor.

[snip of why we'd no longer need telephoto lenses]

> The reason why you don't need it is because you just shoot away covering the
> whole scene you want to shoot, and crop the part of the photo you want
> because you have a gigapixel, or terapixel body.

At some point you will become limited by photon noise. After that you
don't want to make your sensing elements any smaller. Compare the noise
levels of an average point-n-shoot to those of a DSLR when both cameras
are at ISO 400 for an example of how smaller sensing elements affect
noise levels in the final image.

Since there is a minimum desireable limit for the size of a sensing
element--beyond which noise becomes impractical to deal with--there
is a limit as to how fine a detail the camera can resolve at the
sensor.

Let's assume for argument's sake that the limit is 1.0 microns for the
size of a sensing element--that's about 1/9 the size of some of the
elements used in some point-n-shoots. A square 1 gigapixel sensor would
measure about 3.1 centimeters across... not too bad in terms of size.
A square 1 terapixel sensor would measure about 1 meter across... not
something you want to use for a walk around town.

You would need sensor elements that are 0.036 microns across--less than
1/10 the size of the shortest light wavelengths they're supposed to
detect--in order for your 1 terapixel sensor to fit comfortably in a
35mm camera body.

> Scary isn't it! Not because of the fact that this future technology is a
> possibility, but because photography will no longer be fun!

Where are these facts of which you speak? I think you may have a lot

BJJB

Skinner1@hotmail.com
Guest
Posts: n/a

 11-17-2004
On Tue, 16 Nov 2004 23:25:47 -0000, "Matt" <(E-Mail Removed)> wrote:

<SNIP>

Very interesting observation.

And very plausable at some point in the future.

One thing I am waiting on is a camera that will mimic the performance
of different traditional film types at the touch of a button.

Michael A. Covington
Guest
Posts: n/a

 11-17-2004

<(E-Mail Removed)> wrote in message
news:(E-Mail Removed)...
> On Tue, 16 Nov 2004 23:25:47 -0000, "Matt" <(E-Mail Removed)> wrote:
>
> <SNIP>
>
> Very interesting observation.
>
> And very plausable at some point in the future.

Only when they invent light that isn't made of waves or photons.

=?ISO-8859-1?Q?BenOne=A9?=
Guest
Posts: n/a

 11-17-2004
Matt wrote:

> Well, Karl, don't get me started on medium/large format of the future.
>
> The digital equivalent will be on satellites. This means that it will be
> possible to take photos of the world, and crop to the part you want!
>
> Just a thought, but is there any reason why this technology doesn't already
> exist, and is already being used by the government?

Glass just isn't up to the task.

--
Ben Thomas
Opinions, conclusions, and other information in this message that do not
relate to the official business of my firm shall be understood as neither
given nor endorsed by it.

Crownfield
Guest
Posts: n/a

 11-17-2004
http://www.velocityreviews.com/forums/(E-Mail Removed) wrote:
>
> On Tue, 16 Nov 2004 23:25:47 -0000, "Matt" <(E-Mail Removed)> wrote:
>
> <SNIP>
>
> Very interesting observation.
>
> And very plausable at some point in the future.
>
> One thing I am waiting on is a camera that will mimic the performance
> of different traditional film types at the touch of a button.

you could have 4 rpogrammable shutter release buttons,
one for each style:
ektachrome, velvia, ...

Michael A. Covington
Guest
Posts: n/a

 11-17-2004

"Crownfield" <(E-Mail Removed)> wrote in message
news:(E-Mail Removed)...
> (E-Mail Removed) wrote:

>> One thing I am waiting on is a camera that will mimic the performance
>> of different traditional film types at the touch of a button.

>
> you could have 4 rpogrammable shutter release buttons,
> one for each style:
> ektachrome, velvia, ...

How about Parameter 1 and Parameter 2 on the Canon Digital Rebel? Higher
contrast and lower contrast respectively (rather like Velvia vs. Kodachrome
64). And also user-settable parameters.

As I understand it, all of these have more dynamic range than film does.
There would be little point in throttling it down to exactly match film.