Another nail in the view camera coffin?

On Sun, 01 Aug 2004 12:57:58 -0500, Leonard Evens
wrote:


It can't be accomplished in a view camera without using digital
techniques. But who said you can't use a view camera and then also
digitally edit it? That way you would have the best of both worlds.
In fact, some of us do that regularly.


What a concept. Wish I'd thought of that!

But... would Stacey approve?


rafe b.
http://www.terrapinphoto.com

After I wrote my long explanation, I drew some pictures and realized the
distinction I was making and my example were wrong. It is always a
mistake to try to do these things in your head. So I will have to eat
crow.

What I said about line of sight was nonsense. The only thing that
determines three dimensional relations is the point of view.

However, there is still a kernel of truth in what I was saying which
exhibits the difference between correcting in the camera and correcting
digitally after exposure. I was basing my conclusion on my experience
as a photographer rather than my analytic training as a mathematician.

Let me try to explain with an example. If I'm still missing something,
perhaps someone can correct me.

Suppose you want to take a picture of a vertical facade without having
the verticals converge. Suppose also you have a camera without
rise/fall capability, but with both front and rear standard center
tilts. Then one well known technique is the following. Pick a vantage
point from which you can get the whole facade in the frame if you tilt
the camera up at an appropriate angle. Now tilt both standards to
bring them back to vertical. Suppose for the sake of the argument that
you tilt the camera centered on the lens. Then this doesn't change the
point of view, and except for the mechanics it yields exactly the same
effect as a rear drop. Verticals in the facade would now be parallel in
the image. But the result of the rear center tilt would be a
persepctive transformation.

So, as Brian claims, you could do the same thing by not tilting the back
and then correcting digitally with a perspective transformation.

There is one problem though.

The argument would be correct if the film frame were infinite (or just
very large). But of course in that case you wouldn't need to point the
camera up to preserve verticals. Now take into consideration the fact
that the film frame is of a fixed size. You have tilted the camera up
so that the facade fits exactly in this fixed frame. Now visualize what
happens as you tilt the film frame about its center to make it vertical.
You change what is contained within the frame, so you no longer have
captured the whole facade, which was the point of it all.

Now work this backward. Suppose the rear standard is vertical and the
frame shows the entire facade. If you get it exactly right, the center
of the frame will image the center of the facade. Now tilt the rear
standard about its center so that it ends up perpendicular to the line
from center image to center facade. You will see again that what is
contained in the frame shifts.

With a view camera you could correct such changes with a rise or fall,
but with a "fixed" camera, you can't. Its lens axis will always be
centered on its frame. So the only way to get the whole image in the
frame is to change the position of the camera, i.e. to change the point
of view, and that will change perspective relations in the way the three
dimensional geometry is recorded in the film. In other words, you can't
reproduce digitally exactly what you get from a specific point of view
by a rise, fall, or shift if you take into account the fixed film size
and the need to change the point of view.

The question then is how much of a change this makes in practical
situations. I've done one calculation, using some simple trigonometry,
which assumes the camera is at ground level and far enough away that the
film distance is very close to the focal length. It is now about 1:30
am, and I got up at about 4am yesterday morning, so I could easily have
made a mistake. But I found that with a 150 mm lens and a 4 x 5 frame
in portrait orientation, you would have to move about 16 percent closer
to a vertical facade in order to get the whole facade in the frame than
you would have to be if you used a front rise or rear fall. That might
be a rather significant difference in point of view and hence in just
what was visible in the image in addition to the facade. With a shorter
focal length lens the percentage difference would be greater.

My argument depends on using the same focal length. You can reproduce
digitally the effect of a rise/fall/shift from the same point of view if
you choose a shorter focal length and crop. But we already know that we
can do that without even pointing the camera up by choosing a short
enough focal length, and my comments about that are still valid.

I've actually learned something from this discussion despite having let
myself get carried away with a fallacious argument.

As has also been noted, there are other advantages to making the
correction in the camera. The main advantage is that it can be tricky
maintaining the horizontal to vertical aspect ratio of the subject by
applying perspective transformations. You can certainly do it, but you
need some additional information to the geometry present in the image.
For example, I believe Panorama Tools requires you to enter the angle
which the camera is tilted up. Doing it naively in Photoshop will
result in an image which is too short for its width. If you do it in
the camera, you get it right by the laws of optics, and you don't have
to fiddle. Another advantage is that each time you do a transformation,
you degrade the image a bit. The more radical the transformation, the
greater the degradation. Finally, in pointing the camera up, in some
circumstances you could have a problem getting all of a building facade
in focus, while with a rise or fall that is seldom a problem.

After I wrote my long explanation, I drew some pictures and realized the
distinction I was making and my example were wrong. It is always a
mistake to try to do these things in your head. So I will have to eat
crow.

What I said about line of sight was nonsense. The only thing that
determines three dimensional relations is the point of view.

However, there is still a kernel of truth in what I was saying which
exhibits the difference between correcting in the camera and correcting
digitally after exposure. I was basing my conclusion on my experience
as a photographer rather than my analytic training as a mathematician.

Let me try to explain with an example. If I'm still missing something,
perhaps someone can correct me.

Suppose you want to take a picture of a vertical facade without having
the verticals converge. Suppose also you have a camera without
rise/fall capability, but with both front and rear standard center
tilts. Then one well known technique is the following. Pick a vantage
point from which you can get the whole facade in the frame if you tilt
the camera up at an appropriate angle. Now tilt both standards to
bring them back to vertical. Suppose for the sake of the argument that
you tilt the camera centered on the lens. Then this doesn't change the
point of view, and except for the mechanics it yields exactly the same
effect as a rear drop. Verticals in the facade would now be parallel in
the image. But the result of the rear center tilt would be a
persepctive transformation.

So, as Brian claims, you could do the same thing by not tilting the back
and then correcting digitally with a perspective transformation.

There is one problem though.

The argument would be correct if the film frame were infinite (or just
very large). But of course in that case you wouldn't need to point the
camera up to preserve verticals. Now take into consideration the fact
that the film frame is of a fixed size. You have tilted the camera up
so that the facade fits exactly in this fixed frame. Now visualize what
happens as you tilt the film frame about its center to make it vertical.
You change what is contained within the frame, so you no longer have
captured the whole facade, which was the point of it all.

Now work this backward. Suppose the rear standard is vertical and the
frame shows the entire facade. If you get it exactly right, the center
of the frame will image the center of the facade. Now tilt the rear
standard about its center so that it ends up perpendicular to the line
from center image to center facade. You will see again that what is
contained in the frame shifts.

With a view camera you could correct such changes with a rise or fall,
but with a "fixed" camera, you can't. Its lens axis will always be
centered on its frame. So the only way to get the whole image in the
frame is to change the position of the camera, i.e. to change the point
of view, and that will change perspective relations in the way the three
dimensional geometry is recorded in the film. In other words, you can't
reproduce digitally exactly what you get from a specific point of view
by a rise, fall, or shift if you take into account the fixed film size
and the need to change the point of view.

The question then is how much of a change this makes in practical
situations. I've done one calculation, using some simple trigonometry,
which assumes the camera is at ground level and far enough away that the
film distance is very close to the focal length. It is now about 1:30
am, and I got up at about 4am yesterday morning, so I could easily have
made a mistake. But I found that with a 150 mm lens and a 4 x 5 frame
in portrait orientation, you would have to move about 16 percent closer
to a vertical facade in order to get the whole facade in the frame than
you would have to be if you used a front rise or rear fall. That might
be a rather significant difference in point of view and hence in just
what was visible in the image in addition to the facade. With a shorter
focal length lens the percentage difference would be greater.

My argument depends on using the same focal length. You can reproduce
digitally the effect of a rise/fall/shift from the same point of view if
you choose a shorter focal length and crop. But we already know that we
can do that without even pointing the camera up by choosing a short
enough focal length, and my comments about that are still valid.

I've actually learned something from this discussion despite having let
myself get carried away with a fallacious argument.

As has also been noted, there are other advantages to making the
correction in the camera. The main advantage is that it can be tricky
maintaining the horizontal to vertical aspect ratio of the subject by
applying perspective transformations. You can certainly do it, but you
need some additional information to the geometry present in the image.
For example, I believe Panorama Tools requires you to enter the angle
which the camera is tilted up. Doing it naively in Photoshop will
result in an image which is too short for its width. If you do it in
the camera, you get it right by the laws of optics, and you don't have
to fiddle. Another advantage is that each time you do a transformation,
you degrade the image a bit. The more radical the transformation, the
greater the degradation. Finally, in pointing the camera up, in some
circumstances you could have a problem getting all of a building facade
in focus, while with a rise or fall that is seldom a problem.

Quoth Leonard Evens :
| brian wrote:
....
| I agree that you can't change line of sight, since that is fixed by
| the position of the entrance pupil of the lens. However, a view
| camera can't change the line of sight any more than a software
| transformation. So what's the point here??
|
| This is the point. With any camera you CHOOSE the line of sight based
| on what you are trying to accomplish. With the fixed camera, you would
| usually have to choose a DIFFERENT line of sight than you would with a
| view camera if you have in mind post exposure digital manipulation.
| Once it is chosen, you can't change it by a plane projective
| transformation of the image.
|
| Consider the typical example of trying to take a picture of a building
| and avoiding having the sides converge vertically. To do that with a
| "fixed" camera, you point it up and then transform the image digitally
| (or optically in an enlarger) so that the sides are parallel. To do it
| with a view camera, you use a rise without changing the line of sight.
| So in the two cases, you have different lines of sight, which result in
| different (three dimensional) relations among elements of the image.

That's an odd way to think of "line of sight", to me. I'd say
you raise the line of sight with a front rise movement. More
about that below.

| All I'm saying is that once you fix the position of the lens, then you
| can play all the shifting and tilting games you want with the rear
| standard, and still be able to duplicate the resulting geometrical
| effects in software.
|
| No. Consider the following example.
|
| Suppose you are taking a picture of a building facade and you don't want
| the sides of the building to appear in the picture, but you can't place
| the camera centered on the building because something is in the way.
| With a view camera, you would place the camera off to one side, with the
| lens axis still perpendicular to the building facade, and use a
| horizontal shift. This will have absolutely no effect on the relations
| of the elements of the facade to one another.
|
| Now suppose you want to do the same thing with a camera without shifts.
| You would have to move to the same location and point the camera so it
| makes an angle with the building facade. Now most likely you will have
| a picture with the front and one side showing and the top and bottom of
| each converging to vanishing points. You can now digitally correct the
| converging horizontals of the facade so they are parallel, but you can't
| get rid of the image of the side by a projective transformation.

That's really not right. For another thought experiment, let your
man with the view camera stay where he is, center the standards
and adjust the camera so that its body points directly at the subject,
like the fixed camera. At this point the two cameras presumably are
equivalent.

Now let him rotate the front and rear standards so they are parallel
with the building, as they were in the shift configuration. During
this procedure, does a side wall disappear from view? No! This is
a rear swing movement that changes size relationships and shaves only
a small fraction of an inch off the image width.

At this point, the camera configuration is the same as it was with
the shift - same rendition, same focus.


Shift can be simply transformed into front and rear swing, and with
that rear swing we are back to projective transformation. (Plus
the focus effect of the front swing, which you can have if I get
to cover the same field of view with a lens a third or fourth the
focal length.) Likewise, front rise can be precisely transformed
into front and rear tilt.

Donn

Quoth Leonard Evens :
| brian wrote:
....
| I agree that you can't change line of sight, since that is fixed by
| the position of the entrance pupil of the lens. However, a view
| camera can't change the line of sight any more than a software
| transformation. So what's the point here??
|
| This is the point. With any camera you CHOOSE the line of sight based
| on what you are trying to accomplish. With the fixed camera, you would
| usually have to choose a DIFFERENT line of sight than you would with a
| view camera if you have in mind post exposure digital manipulation.
| Once it is chosen, you can't change it by a plane projective
| transformation of the image.
|
| Consider the typical example of trying to take a picture of a building
| and avoiding having the sides converge vertically. To do that with a
| "fixed" camera, you point it up and then transform the image digitally
| (or optically in an enlarger) so that the sides are parallel. To do it
| with a view camera, you use a rise without changing the line of sight.
| So in the two cases, you have different lines of sight, which result in
| different (three dimensional) relations among elements of the image.

That's an odd way to think of "line of sight", to me. I'd say
you raise the line of sight with a front rise movement. More
about that below.

| All I'm saying is that once you fix the position of the lens, then you
| can play all the shifting and tilting games you want with the rear
| standard, and still be able to duplicate the resulting geometrical
| effects in software.
|
| No. Consider the following example.
|
| Suppose you are taking a picture of a building facade and you don't want
| the sides of the building to appear in the picture, but you can't place
| the camera centered on the building because something is in the way.
| With a view camera, you would place the camera off to one side, with the
| lens axis still perpendicular to the building facade, and use a
| horizontal shift. This will have absolutely no effect on the relations
| of the elements of the facade to one another.
|
| Now suppose you want to do the same thing with a camera without shifts.
| You would have to move to the same location and point the camera so it
| makes an angle with the building facade. Now most likely you will have
| a picture with the front and one side showing and the top and bottom of
| each converging to vanishing points. You can now digitally correct the
| converging horizontals of the facade so they are parallel, but you can't
| get rid of the image of the side by a projective transformation.

That's really not right. For another thought experiment, let your
man with the view camera stay where he is, center the standards
and adjust the camera so that its body points directly at the subject,
like the fixed camera. At this point the two cameras presumably are
equivalent.

Now let him rotate the front and rear standards so they are parallel
with the building, as they were in the shift configuration. During
this procedure, does a side wall disappear from view? No! This is
a rear swing movement that changes size relationships and shaves only
a small fraction of an inch off the image width.

At this point, the camera configuration is the same as it was with
the shift - same rendition, same focus.


Shift can be simply transformed into front and rear swing, and with
that rear swing we are back to projective transformation. (Plus
the focus effect of the front swing, which you can have if I get
to cover the same field of view with a lens a third or fourth the
focal length.) Likewise, front rise can be precisely transformed
into front and rear tilt.

Donn

Leonard Evens wrote:
After I wrote my long explanation, I drew some pictures and realized the
snip...

I always appreciate Mr. Evens' knowledge and enthusiasm for math and
large format photography. If you have not already done so, check out
Brian Caldwell's web site (URL in his post). He has done a fair amount
of work with image transformation and many of the technical issues
discussed in this thread.

D. Poinsett

Leonard Evens wrote:
After I wrote my long explanation, I drew some pictures and realized the
snip...

I always appreciate Mr. Evens' knowledge and enthusiasm for math and
large format photography. If you have not already done so, check out
Brian Caldwell's web site (URL in his post). He has done a fair amount
of work with image transformation and many of the technical issues
discussed in this thread.

D. Poinsett

Robert Feinman wrote in message . ..
Leaving aside film size, the two features that view cameras still
have over other formats are the ability to adjust perspective and
the plane of focus.
By using a digital editor one can generate the same perspective
effects from an image taken with a conventional camera afterwards.

I've been playing with this feature in Photoshop and have put up
an additional tip about this on my web site. This one shows the
creative uses the extreme perspective adjustments can yield.
Just follow the tips link on my home page, if you are interested.

I still haven't solved the plane of focus problem, however...

Some interesting responses to your post, so I'll only add my personal
view. I don't do photography to sit in front of a computer. I do it
to be there, to capture the best image I see. I'm just starting to
learn to large-format from 35 years of 35mm photography, using two PC
lenses in my system. I plan to enjoy the time it takes in
large-format. The rest just is, to produce the best image from the
field work, and not have to rely on PS to do what I can do in the
field.

In article , Stacey
wrote:

dr bob wrote:


Here's another anecdote: My 4x5 Speed Graphic was setup for a landscape.
An individual appeared and a casual conversation began. I'm sure he/she
wanted to "see" whatever I was photographing. "Do they still make those
old
cameras?"

Snip

My last experience was this guy came over and said "You know professionals
use digital cameras now." My responce was 'I'm not very professional'.

Assholes are everywhere, always trying to tell someone something they
think makes them smarter than you, fact is they are just insecure enough
about their own real knowledge to try and make you uncomfortable about
what your doing.

If you have better pictures than any they have seen they really can't
say much.
--
To win one hundred victories in one hundred battles is not the
measure of skill. To subdue the enemy without fighting is the
measure of skill. Sun Tzu

In article , Stacey
wrote:

dr bob wrote:


Here's another anecdote: My 4x5 Speed Graphic was setup for a landscape.
An individual appeared and a casual conversation began. I'm sure he/she
wanted to "see" whatever I was photographing. "Do they still make those
old
cameras?"

Snip

My last experience was this guy came over and said "You know professionals
use digital cameras now." My responce was 'I'm not very professional'.

Assholes are everywhere, always trying to tell someone something they
think makes them smarter than you, fact is they are just insecure enough
about their own real knowledge to try and make you uncomfortable about
what your doing.

If you have better pictures than any they have seen they really can't
say much.
--
To win one hundred victories in one hundred battles is not the
measure of skill. To subdue the enemy without fighting is the
measure of skill. Sun Tzu