Hacker News Reader

> When people talk about how a focal-length "compresses" or "exaggerates" an image, what they're really talking about is perspective. Each field of view carries with it its own perspective.

Completely agree. What is happening when the focal length of a camera is changed is that the distance of the vanishing points relative to the scene is changing.

I teach perspective in my drawing class. Few subjects are as difficult to convey. Using 3d animations has helped a lot…

https://rmit.instructure.com/courses/87565/pages/perspective...

The "dolly zoom" is a very effective way to convey this imo.

https://filmschoolrejects.com/wp-content/uploads/2021/01/Jaw...

https://filmschoolrejects.com/wp-content/uploads/2021/01/Fel...

https://screenqueens.files.wordpress.com/2017/08/polt.gif

https://screen-queens.com/2017/08/24/how-the-dolly-zoom-chan...

There's also some rather neat series that play with perspective. Don't have any at hand but google:

https://photography.tutsplus.com/tutorials/exploring-how-foc...

https://static.wixstatic.com/media/7d67c8_1d09b657f6a04e6fb8...

The way I've always explained it to people as a photographer: your position with regard to the subject determines the perspective, which is the relationship between objects in the composition. Your choice of focal length determines what field of view/image area of that perspective that you want to capture.

And yes, to agree with a parent comment, this is something that zoom lenses have rotted the brains of many photographers on. People stand wherever is convenient and then zoom to fit the subjects, rather than thinking first about how they want the elements to be placed within the image. You should move to where the elements have the proper relationship, then figure out your field of view from there.

To use the Jaws example: if you want the tent in the frame, you need to be close and go wide. If you want to exclude the tent from the frame, you go farther away and zoom in. As a photographer, you need to make that decision about your composition first, and then choose the position that gives you that, and then select your field of view last.

It's hugely beneficial imo to work with a 3-lens kit or something similar (eg 28mm, 50mm, 105mm) for at least a little while, to really force you to think about what you want from a particular exposure/composition.

A useful implication of this: you get the exact same perspective from taking a picture with a wide angle lens and cropping it to your desired composition, as if you had taken the same photo with a long lens and not cropped it at all. You just get less resolution - so it's obviously desirable to use a longer lens if you have one - but the perspective is the same either way. Similarly - if your lens isn't wide enough, you can stitch together multiple exposures and it will be the same as if you had used a wider lens. Perspective is completely independent of focal length of the lens, it's purely a property that emerges from the relative positions of the subject and photographer. All the lens is doing is picking a "crop".

(to get really really tangential: this is one of the problems I have with "photojournalism" as a field of photography and photojournalism contests in particular, because it completely forbids things like cropping as somehow tampering with the essence of an image. But cropping is a natural part of composition, you do it with your lens even if you are not doing it in photoshop. The photographer (if they're good) chose to exclude the mother idly smoking a cig as she waits for the bus and compose the screaming baby in the carriage instead. Pretending like there is a "platonic truth" to an image because it "wasn't edited" despite being deliberately constructed by a photographer with an agenda (towards drama and impact, if nothing else) and editorial and artistic license is a microcosm of the way modern media feigns complete neutrality and unbiasedness by "just reporting facts" while completely ignoring the agendas of those who choose what facts are reported and the framing used. Anyway, those "photojournalism contests" should really be called "straight-out-of-camera" contests, and even then there is still processing happening. There is no platonic "right" way to process a RAW into a JPG, even if it's the camera doing it itself. But "straight-out-of-camera awards wouldn't be something that sounds important enough to give a bunch of money and adulation over :])

> There is no platonic "right" way to process a RAW into a JPG,

This. So much this. RAW to JPEG conversion is an opinionated act. Just look at the different way phones take pictures. And this isn’t even getting into lens and sensor design.

Google explicitly does a lot of computational photography. That what makes Pixel phones produce pretty decent images out of relatively humble hardware.

Yes, this has gotten even wackier in the era of "neural image enhancement". A phone pic isn't a picture, it's a "DLSS of a picture", or "deep dream of a picture".

And to agree with grandparent - developing a photo has never been a neutral act, whether it's digital or film. The choice of developer and processing has enormous impact on an image. It can increase or reduce contrast, grain, sharpness, etc which can completely change the feel of an image. And that's not just developer choice and times/temps, but things like selenium/chromium/mercury intensification, staining developers (that modulate image contrast/etc) like pyrocat, etc.

All of modern photoshopping evolved from analog techniques - they call it an "unsharp mask" because photographers would actually make a mask using a blurred copy of the image, which cancels out some of the signal and increases the sharpness of the remaining edges. And even before the negative is developed, you can do all kinds of things to control the feel of the image.

In fact for about the first 1/3 of photography's existence it was practically necessary to do intensification - you will find tons of references to it in older guides.

A Chrome Experiment from back in the day that I remember fondly:

https://experiments.withgoogle.com/dollyzoomjs

(then found at https://web.archive.org/web/20140423002845if_/https://www.ch...)

Much love to this demo. Very useful.

> When people talk about how a focal-length "compresses" or "exaggerates" an image, what they're really talking about is perspective. Each field of view carries with it its own perspective.

This is kind of a misconception because what creates perspective distortion is not a given field of view, but the distance to the subject. That's why short (50-100 mm) macro lenses have a lot of perspective despite a clearly tele angle of view; you get very close to the subject. Someone using a 55 mm macro lens for photography of small products tends to be pretty obvious because of the strong perspective, things look distorted and "bulging".

Of course subject size / angle of view => distance so it still kinda works.

Yeah, I was going to point out that out when he talked about the portrait. It's not really about the focal length but about the distance you normally look at people (although your brain does a lot of compensation for wonky perspectives). Focal length doesn't change the perspective, but it just crops the image. Your feet changes the perspective.

And, if you used an infinite zoom macro lens on a subject infinitely far away, it would not have perspective at all but instead appear as an isometric view. Where does one find an infinitely large lens? Don't worry, you can't afford it anyway. It costs infinite money.

Jokes aside, the math on this really works out. Orthogonal projection is the limit as d tends to infinity of perspective projection.

That's sort of what a telecentric lens does.

https://www.lockhaven.edu/~dsimanek/3d/telecent.htm

https://www.youtube.com/watch?v=8xP4lr1ojTo

And you can go further: the hypercentric lens.

https://www.youtube.com/watch?v=iJ4yL6kaV1A

> Where does one find an infinitely large lens? Don't worry, you can't afford it anyway. It costs infinite money.

NRO: haha ok but anyway, where can we get one

Right, this is how I understand it too. Is it not true if you took two photos of the same subject, with say a 24mmm lens and 70mm lens, and the then cropped and enlarged the image from the 24mm lens shot, you'd end up with the same image with the same perspective?

If the two images are made with the same distance between subject and camera, then yes.

Tank Man is not a good example here, as it is not just one image. People might remember [1], but never seen [2] or not even be aware that there is a full video of the event [3] as well. And of course depending on where you viewed it, it might have been cropped, photoshopped or taken from the video.

[1] https://img.timeinc.net/time/images/covers/asia/2001/2001011...

[2] https://miro.medium.com/max/1024/1*yMZA-hOHadTkNrGrSKriBw.jp...

[3] https://www.youtube.com/watch?v=qq8zFLIftGk

In an important sense, all photography is compatible with linear perspective. Most lenses aim to closely emulate rectilinear perspective, and those that don't are still constrained by the location of the lens in space.

A pinhole camera can see certain points in any scene, and not others. These points are the same points as a traditional linear perspective rendering with the same viewpoint would represent. Now, lenses do sometimes distort things, but they can't see areas of the scene that are occluded: light travels in straight lines. IMO there is something really straightforward about this situation. Any image is constrained by the lines of sight from the viewpoint, and linear perspective is just a projection in which to display this data, which is (to a very close approximation) the same, however you capture the image. So linear perspective, from this point of view, has little or nothing to do with lens-based media. It's a total commonplace of image formation and it doesn't have any consequences for how you use the camera to portray depth. Those decisions are unrelated to perspective. The fact that the lens can only see non-occluded points in the scene is the determining factor.

The article is about more creative problems that arise when an artist interprets a scene, unconstrained by geometry in a way cameras never are. "This paper points out that artists almost never use “correct” linear perspective"

No, the Tank Man picture isn't surprising at all. The defining property of that image is that he is small, but willing to stand up to those tanks. My memory was correct -- it is a picture of a very small object standing in front of several extremely large objects.