Perspective control using shifts

A view camera has a lens that produces an image circle much larger then
is needed for the film size. When shifts are used you are selecting
what part of this circle gets captured to film. Shifts do not change
the perspective of the image, only the aiming of the lens changes that.
If we wish to have a photograph that does not have converging vertical
lines then the lens needs to be pointed horizontally, we then shift the
relationship between the lens and film to frame the shot.

This image shows what this would look like if you could see the whole
image circle.
http://www.sewcon.com/temp/shift.jpg

The rectangle represents one possible position for the film. Note that
the camera is pretty much pointing towards the horizon, if I wanted
just a bit of convergence it could be pointed up just a bit.

It would be possible to build a camera that has no shifts but can
correct for perspective, simply us a piece of film that is large enough
to cover the image circle and then crop what post of the image you
want. The shift part of the camera makes far better use of the film
and allows for a much smaller camera.

Scott

I don't understand your argument. Covering the image circle is what
35mm does. You're suggesting just use a short enough focal length lens
to capture the whole subject and crop it? Why not just use a 35mm
film/digital tilt-shift lens for the right focal length and correct for
convergence/divergence in the field? And you don't need a LF camera.

"Scott W" wrote in message
ups.com...
A view camera has a lens that produces an image circle much larger then
is needed for the film size. When shifts are used you are selecting
what part of this circle gets captured to film. Shifts do not change
the perspective of the image, only the aiming of the lens changes that.
If we wish to have a photograph that does not have converging vertical
lines then the lens needs to be pointed horizontally, we then shift the
relationship between the lens and film to frame the shot.

Strictly, that may seem true if you are used to cameras where the
relationship between film and lens is fixed, but actually what prevents
converging verticals is not where the lens is pointed: it's having the film
plane vertical. To be precise, for a shape - say, the rectangle of a
building - to appear on the film as the same shape, with no lines that
converge, angles changed, circles turned to elipses, or similar, the film
plane must be parallel to the plane that that shape lies in. So verticals
don't converge if the film back is vertical too.

Shifting the lens (or the film relative to the lens) moves which bit of the
image circle falls on the film and so allows you, given enough coverage, to
keep the film parallel to the plane you are interested in but record a part
of that plane that isn't directly in front of the camera: the classic
example is shifting the lens up relative to the film in order to include the
top of a tall building without tilting the film plane because to do so would
make the verticals converge.

Tilting the lens - 'aiming' it as you say - does not alter the geometry of
shapes.

So (to simplify a lot) if the film is vertical, and so parallel to a tall
building, you can include the top of the building only by shifting the lens
up (or the film down) so that the line ('ray') from the bottom of the film
through the optical centre of the lens goes on to include the top of the
building. Whether the lens is angled up or down makes no difference to
which part of its image circle you are recording or to the shapes of the
objects projected. What it does do is alter which part of the projected
image cone the film intersects and so where the plane of sharpest focus lies
through the field of view.

ie. - shifts affect geometry, lens tilts affect focus.

It is only when you tilt the film plane that you (significantly) affect both
geometry and focus.

To give an example: I am photographing a pool table from a position a foot
above the end cushion. I can shift the lens down a little to improve the
amount of the near end of the table that is in my field of view, without
altering the geometry of what I see. I can also use tilts to make the plane
of sharpest focus lie parallel to the table instead of perpendicular to it.
If I tilt the lens forward I won't alter the geometry of the table in the
image, but I can get my plane of focus aligned parallel to the table and so
get all the balls easily within the DoF - however, such a tilt might, even
with the downward shift ('fall') I have already applied to the lens, mean
that the lens can no longer project a complete image onto the film because
the film now lies partly outside its circle of coverage. The alternative is
to leave the lens untilted and instead tilt the film backward a bit. This
can again align the plane of focus where I want it, and can do so without
moving the film outside the lens' circle of coverage. However, it will
alter the apparent geometry of the image, making the near end of the table,
and the balls on it, seem relatively larger than the far end.

(If we think about it, this is obvious: the image of the far end of the
table is projected at the bottom of the film plane: by tilting the film
backward I have made the top of the film plane relatively further from the
lens than the bottom, so the far end of the table is magnified less, and the
near end more, in the projected image. For the same reason, the near end of
the table will be a little less fully exposed than the far end, because the
light has effectively been projected down a greater bellows extension, where
it falls off according to the inverse square law.)

In practice, for such a shot, I might want a bit of extra emphasis on the
near end of the table and be happy to exaggerate the size of the balls
there, so I might us some back (film plane) tilt, but I might use front
(lens) tilt as well if using all back tilt made the effect more pronounced
than I wanted. I could even 'over-tilt' the film to give extreme effects,
and then correct the plane of focus with a little back tilt on the lens -
any combination that gives the image I want is possible, so long as the film
stays within the circle of projection of the lens.

I hope that rather simplistic explanation makes sense: I just wanted to
emphasise that it is not really where we 'point the lens' that controls the
geometry of objects in the image, it is where we place the film that does
that.


Peter

Bandicoot wrote:
"Scott W" wrote in message
ups.com...

A view camera has a lens that produces an image circle much larger then
is needed for the film size. When shifts are used you are selecting
what part of this circle gets captured to film. Shifts do not change
the perspective of the image, only the aiming of the lens changes that.
If we wish to have a photograph that does not have converging vertical
lines then the lens needs to be pointed horizontally, we then shift the
relationship between the lens and film to frame the shot.

. . . . . . . . . .
I hope that rather simplistic explanation makes sense: I just wanted to
emphasise that it is not really where we 'point the lens' that controls the
geometry of objects in the image, it is where we place the film that does
that.


Peter



Just to add to that great explanation, here is a shot using a simple
shift lens on 35 mm:

http://www.allgstudio.com/gallery/SFX/balls_SFX.jpg

These are all the same round shape, but the camera position and a shift
of the lens in a diagonal direction causes the shapes in the background
to go more oblong, and appear slightly tilted. The shape in the
foreground is actually very close to how these appear at that location.

Front shift and rear shift on a view camera can function somewhat
differently than a shift lens on smaller formats. I think part of that
is altering the lens alignment to the subject/scene, while in other
instances is changing the alignment of the film plane to the subject/scene.

Ciao!

Gordon Moat
A G Studio
http://www.allgstudio.com

I think you missed something. The image on the film resulting
from a shift is from a different part of the image circle.
Just using a larger sheet of film with a wider lens doesn't change that
characteristic.
The result is that nothing has changed.

Collin

"Scott W" wrote in message
ups.com...

http://www.sewcon.com/temp/shift.jpg


A Volkswagen and a Mini Cooper! You're obviously not in GM country!
;-)

--
Regards,
Matt Clara
www.mattclara.com

Hi all,

Could somebody post the formula for the inverse law as it relates to
bellows extension so that I can calculate the increase in f stops and
compensate for the exposure.
My archives got scrambled and I can't find it. Thanks.

Regards,
Bogdan



Bandicoot wrote:

snip

IFor the same reason, the near end of
the table will be a little less fully exposed than the far end, because the
light has effectively been projected down a greater bellows extension, where
it falls off according to the inverse square law.)

snip
--
__________________________________________________ ________________
Bogdan Karasek
Montr‚al, Qu‚bec e-mail:
Canada

"Wovon man nicht sprechen kann, dar�ber muss man schweigen"
"What we cannot speak about we must pass over in silence"
Ludwig Wittgenstein
__________________________________________________ ______________

Scott W wrote:

A view camera has a lens that produces an image circle much larger then
is needed for the film size. When shifts are used you are selecting
what part of this circle gets captured to film. Shifts do not change
the perspective of the image, only the aiming of the lens changes that.

Front shift/lens shift changes perspective. Try this experiment.

Set your camera up with the lens shifted all the way to the left.
Compose a macro image (1:1) of a ruler, say the top 3 inches. Just
behind the ruler, about 2" or so, place a page of text, so that the text
is visible on both sides of the ruler. Use a small f/stop to insure DOF
keeps the text readable.

Make an exposure.

Now without changing anything else, shift the lens all the way to the
right. Make another exposure.

Compare the two exposures, and you'll see the perspective between the
ruler & text has changed. The ruler will appear to have moved in front
of the text.

Rear shift moves around the image circle. Front/lens shift changes
perspective.

Neither shift changes focus.

no_name wrote:
Scott W wrote:

A view camera has a lens that produces an image circle much larger then
is needed for the film size. When shifts are used you are selecting
what part of this circle gets captured to film. Shifts do not change
the perspective of the image, only the aiming of the lens changes that.

Front shift/lens shift changes perspective. Try this experiment.

Set your camera up with the lens shifted all the way to the left.
Compose a macro image (1:1) of a ruler, say the top 3 inches. Just
behind the ruler, about 2" or so, place a page of text, so that the text
is visible on both sides of the ruler. Use a small f/stop to insure DOF
keeps the text readable.

Make an exposure.

Now without changing anything else, shift the lens all the way to the
right. Make another exposure.

Compare the two exposures, and you'll see the perspective between the
ruler & text has changed. The ruler will appear to have moved in front
of the text.

Rear shift moves around the image circle. Front/lens shift changes
perspective.

Front shift changes the point of view and the framing but it does not
change the vanishing points in the image.

Scott

Bogdan Karasek wrote:

Hi all,

Could somebody post the formula for the inverse law as it relates to
bellows extension so that I can calculate the increase in f stops and
compensate for the exposure.
My archives got scrambled and I can't find it. Thanks.

Regards,
Bogdan



Bandicoot wrote:

snip

IFor the same reason, the near end of
the table will be a little less fully exposed than the far end,
because the
light has effectively been projected down a greater bellows extension,
where
it falls off according to the inverse square law.)


snip


BEF = Image Distance^2/focal length^2

(square of image distance divided by square of the focal length)

Stated as 'x'

1x = 0 stops
2x = 1 stops
3x = 1-1/2 stops
4x = 2 stops
8x = 3 stops
16x = 4 stops
32x = 5 stops