Question about rectilinear ultrawide lens ...

Started Jan 18, 2008 | Discussions
kguru Contributing Member • Posts: 518
Question about rectilinear ultrawide lens ...

With a rectilinear ultrawide lens persons standing near the edge look wider than those in the middle. Please explain what technical constraints prevent designers from making these lenses so that a person looks the same width and same height no matter where he stands edge or middle. Thanks.

matthis New Member • Posts: 20
Re: Question about rectilinear ultrawide lens ...

Geometry. Put several people in a row parallel to the sensor plane and it won't matter if they are in the center or near the edge.

DrewE Senior Member • Posts: 2,053
Re: Question about rectilinear ultrawide lens ...

In a word, geometry. It's inherent in a rectilinear projection.

A rectilinear lens, for the purposes of this (attempted) analysis, may be thought of as a point through which the light passes and is projected onto the film/sensor plane. (Or, to put it another way, think of a pinhole camera's pinhole "lens.") The focal length of the lens at the extreme edges of the film/sensor is different than at the center. For a relatively long lens (or a pinhole relatively distant from the film), this difference is miniscule enough to neglect; for a relatively short lens (or pinhole close to the film), however, it becomes significant. It's a similar reason why the sizes of land masses near the equator of a mercator projection map are smaller appearing than those near the poles.

A little thought along these lines will also show why it's not possible to have a 180° or larger field of view with a rectilinear lens.
--
--DrewE

 DrewE's gear list:DrewE's gear list
Pentax K10D Pentax K-5 Pentax smc DA 18-55mm F3.5-5.6 AL Pentax smc DA 40mm F2.8 Limited Voigtlander 58mm F1.4 Nokton SL II +3 more
Ricardo Tomasi Regular Member • Posts: 182
Re: Question about rectilinear ultrawide lens ...

The only constraints for doing it on a lens is... physics.

You can probably "fix" that on software, but you'll end up with a distorted and ruined perspective.

Do you have special eyes that make everything the same size, doesn't matter the distance?

OP kguru Contributing Member • Posts: 518
Re: Question about rectilinear ultrawide lens ...

I appreciate your replies, but there's still something missing.

If due to geometry, does it mean if I take a photo of a tall building, making sure it's paralell with the sensor plane, the top floors will look taller than middle floors? (similar to edge persons fatter than middle persons)

Sean Nelson
Sean Nelson Forum Pro • Posts: 13,173
Re: Question about rectilinear ultrawide lens ...

Let me see if I can explain what an ultra-wide rectlinear lens is doing using some photos...

Here's a picture of my rec room taken with a 15mm SMC-M Pentax lens (on my K100D Super):

For the most part, the lens has rendered straight lines straight. This is what's meant by "rectlinear".

Now lets suppose I didn't have an ultra-wide angle lens. What I could do instead would be to take a panorama of three shots using a narrower lens and then stitch them together. Here's three shots taken in portrait mode with the kit lens at a focal length of 24mm, put together so you can clearly see each indvidual frame:

Notice how the horizontal lines converge in the side photos. If I stitch these three photos together as is, I would end up with a "fisheye" effect. The result would NOT be a rectlinear rendering of the subject.

In order to acheive a rectlinear perspective, an ultra-wide angle lens "distorts" the edges of the picture in order to render all lines straight. It's as if I took the rightmost picture from the panorama and used Photoshop distortion tools to stretch the image from the original (left) image below to the "corrected" version on the right:

In doing this, items toward the edges and corners get distorted. Here's a close-up of the digital clock - you can see that in the rectlinear rendition it doesn't look the way it's supposed to:

This is an unavoidable side-effect of rectlinear rendition over wide angles. The truth of the matter is that a fisheye lens actually renders wide angles more like the eye sees them. The problem is that when you reduce a huge 100-degree-plus fisheye view into a small image viewed at an apparent angle of 30 degrees or so it doesn't look natural because our eyes don't expect to see things look fisheye-like over a small 30-degree field of view.

Rectlinear lenses are wonderful for rendering straight lines straight - but the spatial relationships of the image have to be bent in order to do this.

EricJohnson New Member • Posts: 13
Re: Question about rectilinear ultrawide lens ...

Ok, this makes sense to me. I asked a question a few minutes ago about the purchase of a wide angle lense to take interior home photos. Based on your comments, should I lean toward a rectilinear lense over a "traditional" fisheye so my pix don't look like they were taken through a Coke bottle? Thanks for the advice...EJ

OP kguru Contributing Member • Posts: 518
Re: Question about rectilinear ultrawide lens ...

Sean Nelson wrote:

Let me see if I can explain what an ultra-wide rectlinear lens is
doing using some photos...

Sean, what can I say ... I'm dumbfounded by the extend you go to, to answer a forum question. The anigifs a few days ago was another example.

Thankyou very much. I'm now going to digest your explanation more thoroughly.

Thomas Kachadurian
Thomas Kachadurian Veteran Member • Posts: 3,656
Re: Question about rectilinear ultrawide lens ...

kguru wrote:

I appreciate your replies, but there's still something missing.

If due to geometry, does it mean if I take a photo of a tall
building, making sure it's paralell with the sensor plane, the top
floors will look taller than middle floors? (similar to edge persons
fatter than middle persons)

Yep, that's exactly right. The difference is that we are keyed into the shape of heads and we let it go when it's parts of a building.

that's why I prefer a fish-eye for shooting people.

Tom
--
http://www.kachadurian.com

 Thomas Kachadurian's gear list:Thomas Kachadurian's gear list
Panasonic LX100 Panasonic Lumix DMC-GX8 Panasonic Lumix DC-G9 Panasonic Lumix G 20mm F1.7 ASPH Panasonic Leica Summilux DG 25mm F1.4 +7 more
richardplondon
richardplondon Forum Pro • Posts: 10,893
Re: Question about rectilinear ultrawide lens ...

kguru wrote:

With a rectilinear ultrawide lens persons standing near the edge look
wider than those in the middle. Please explain what technical
constraints prevent designers from making these lenses so that a
person looks the same width and same height no matter where he
stands edge or middle. Thanks.

If you stand next to a straight line painted on the ground, you will see that the line slopes one way when you look to your left, the other way to your right, and runs straight across in the middle. There are no bends in the painted line, yet it really does this, optically speaking.

The change of apparent angle is because of the change of viewing direction .

The fisheye approach acknowledges this (pragmatically), and the rectilinear approach refuses to.

Like all denials of reality, rectilinear / classic perspective involves you in more and more extreme distortions the further you get from the central point. These are the costs of keeping straight lines straight.

It is all convincing and familiar, of course, with normal or moderately wide angle lenses. Even extremely distorted, ultrawide rectilinear images CAN look convincing, but only if viewed in the right way.

a) As part of rectilinear correction, objects have to stretch in the radial direction, more so the further away from the centre they are.

b) Looking at any large flat surface from close to its centre, things drawn on the surface foreshorten in the radial direction, more so the further away from the centre they are.

From the right viewpoint, these two effects (the stretch and the shrink) can cancel each other out, so you don't perceive any distortion. But the image needs to be seen very "large" (IMAX style) or from very close.

For interior shooting of very large angles, you cannot expect your public to undergo such trials. Fisheye can look too weird. But panorama stitching is good at "managing" quite large angles of view, giving you a choice of mathematical projections, and interactive panorama viewers can deliver even a 360 x 180 sphere in full cinematic immersion. It may be worth looking into that; often used in the property business.

RP

 richardplondon's gear list:richardplondon's gear list
Panasonic LX10 Pentax K-5 Sigma 10-20mm F4-5.6 EX DC HSM Pentax smc DA 21mm F3.2 AL Limited Pentax smc DA 70mm F2.4 AL Limited +7 more
alex wetmore Senior Member • Posts: 1,394
Re: Question about rectilinear ultrawide lens ...

One thing to remember is that a fisheye can actually look more natural than a rectilinear in some situations. It took me a long time to accept this.

A moderately wide fisheye (10-17FE on the 17 side) looks pretty similar to human vision. Things in the periphery start to curve, but that is true of my peripheral vision too. At 17mm the 10-17FE is already "wider" than the 16mm end of the 16-45 and getting close to the 14mm rectilinear.

I think that landscapes taken with a moderate fisheye can look more natural. There aren't that many perfectly straight lines in landscapes. In a fisheye things in the corners of the frame aren't being strangely stretched to make them fit.

If Pentax made a 15mm/4 fisheye prime pancake I'd be all over it.

Here is a 17mm example:

13mm:

With my DA 14/2.8 there would have been much more noticable (to me) distortion. In the 13mm example you can see distortion (the roof of the building isn't straight), but it doesn't scream out as much. We tend to look for humans, and the distorted humans would be more distracting.

A fisheye will distort humans into odd curves if they are on the very edge of the frame:

I didn't know what to expect of the fisheye zoom when I first bought it. I bought the 14/2.8 at the same time. The fisheye zoom has become one of my favorite lenses. I use the 14/2.8 too, but not as often.

alex

DrewE Senior Member • Posts: 2,053
Re: Question about rectilinear ultrawide lens ...

For a theoretically perfect rectilinear lens, the top floors will occupy the same space as the middle floors in the resulting image. This is precisely the distortion you see in shots of people, as the top floors are farther away than the middle floors and thus "should" look smaller.
--
--DrewE

 DrewE's gear list:DrewE's gear list
Pentax K10D Pentax K-5 Pentax smc DA 18-55mm F3.5-5.6 AL Pentax smc DA 40mm F2.8 Limited Voigtlander 58mm F1.4 Nokton SL II +3 more
OP kguru Contributing Member • Posts: 518
Re: Question about rectilinear ultrawide lens ...

DrewE wrote:

For a theoretically perfect rectilinear lens, the top floors will
occupy the same space as the middle floors in the resulting image.
This is precisely the distortion you see in shots of people, as the
top floors are farther away than the middle floors and thus "should"
look smaller.
--
--DrewE

If this "distortion" makes things of same dimension occupy same space in the resulting image, it should not make people (talking about a group photo with people standing across on the same focal plane) look fatter near the edge. That's what got me puzzled Drew.

richardplondon
richardplondon Forum Pro • Posts: 10,893
Re: Question about rectilinear ultrawide lens ...

kguru wrote:

DrewE wrote:

For a theoretically perfect rectilinear lens, the top floors will
occupy the same space as the middle floors in the resulting image.
This is precisely the distortion you see in shots of people, as the
top floors are farther away than the middle floors and thus "should"
look smaller.
--
--DrewE

If this "distortion" makes things of same dimension occupy same space
in the resulting image, it should not make people (talking about a
group photo with people standing across on the same focal plane) look
fatter near the edge. That's what got me puzzled Drew.

This distortion (rectilinear correction) does not happen equally in all directions, but only in the radial direction, whatever that might be for each part of the image. Sizes perpendicular to that are not affected.

So an item at the top of the image keeps its width, but changes height. An object at the left of the image keeps its height, but changes width. An object in the corner keeps one diagonal, but changes the other. All round the image, shapes stretch relative to the centre.

The reason for all this stretching, IMO, is to construct an ideal pictorial viewpoint - not to emulate real visual experience. Though our brains don't let us experience literal fisheye vision, either. We don't seem to see "in" any single optical projection.

RP

 richardplondon's gear list:richardplondon's gear list
Panasonic LX10 Pentax K-5 Sigma 10-20mm F4-5.6 EX DC HSM Pentax smc DA 21mm F3.2 AL Limited Pentax smc DA 70mm F2.4 AL Limited +7 more
OP kguru Contributing Member • Posts: 518
Re: Question about rectilinear ultrawide lens ...

richardplondon wrote:

kguru wrote:

DrewE wrote:

For a theoretically perfect rectilinear lens, the top floors will
occupy the same space as the middle floors in the resulting image.
This is precisely the distortion you see in shots of people, as the
top floors are farther away than the middle floors and thus "should"
look smaller.
--
--DrewE

If this "distortion" makes things of same dimension occupy same space
in the resulting image, it should not make people (talking about a
group photo with people standing across on the same focal plane) look
fatter near the edge. That's what got me puzzled Drew.

This distortion (rectilinear correction) does not happen equally in
all directions, but only in the radial direction, whatever that
might be for each part of the image. Sizes perpendicular to that are
not affected.

So an item at the top of the image keeps its width, but changes
height. An object at the left of the image keeps its height, but
changes width. An object in the corner keeps one diagonal, but
changes the other. All round the image, shapes stretch relative to
the centre.

The reason for all this stretching, IMO, is to construct an ideal
pictorial viewpoint - not to emulate real visual experience. Though
our brains don't let us experience literal fisheye vision, either. We
don't seem to see "in" any single optical projection.

RP

Thank you Richard, your explanation makes things very clear.

By the way, can't they design lenses to be free of this stretching without becoming fisheye?

Sean Nelson
Sean Nelson Forum Pro • Posts: 13,173
Re: Question about rectilinear ultrawide lens ...

EricJohnson wrote:

should I lean toward a rectilinear
lense over a "traditional" fisheye so my pix don't look like they
were taken through a Coke bottle?

Absolutely. A rectlinear ultra-wide angle lens is a real estate agent's dream - it can take any old cramped room and make it look spacious. Just make sure there aren't any people in the photos!

Jan Moren Veteran Member • Posts: 3,745
Re: Question about rectilinear ultrawide lens ...

kguru wrote:

Thank you Richard, your explanation makes things very clear.
By the way, can't they design lenses to be free of this stretching
without becoming fisheye?

Nope.

Or rather, there are other possible mappings (used in cartography for instance) but they will look stranger than a fisheye, and as far as I know have never been realized as a physical lens.

OP kguru Contributing Member • Posts: 518
Re: Question about rectilinear ultrawide lens ...

alex wetmore wrote:

One thing to remember is that a fisheye can actually look more
natural than a rectilinear in some situations. It took me a long
time to accept this.

Thank you Alex.

What you said there is true; if there aren't straight lines near the edge the fisheye effect doesn't really scream out at you; even at 10mm as in this pic:

Jonson PL Veteran Member • Posts: 3,600
Re: Question about rectilinear ultrawide lens ...

Thanks a lot Sean, for your lengthy explanation, very helpful. (and others).

Never really quite grasped the concept between rectliniear and fisheye. Though rectliniear was always to prefer

-- hide signature --

Kind regards
Sune

Kuala Lumpur, Pangkor Laut, Bangkok imagery :
http://forums.dpreview.com/forums/readflat.asp?forum=1036&thread=26227009

jaad75 Contributing Member • Posts: 601
Re: Question about rectilinear ultrawide lens ...

alex wetmore wrote:

At 17mm the 10-17FE is
already "wider" than the 16mm end of the 16-45 and getting close to
the 14mm rectilinear.

Actually it is close to 12mm rectalinear. DA12-24 has 99 degrees FOV, it's the same as DA10-17 at longer end...

-- hide signature --
Keyboard shortcuts:
FForum MMy threads