[SamBam]

> By legibility I mean that the Piranesi distortion is easier to read, and that to anyone unfamiliar with photographs, that hasn’t grown up with TV, photographs and magazines, the Piranesi version would look much better, and the true perspective would look rather odd.

I think this is a key point that people are often unaware of. For instance, whenever I see a modern artist who has painted photorealistic art, I read discussion about how amazing it is that this artist can paint something so perfectly life-like, and the Renaissance artists were clearly inferior, since they couldn't paint with such skill.

Besides the obvious discussion about the purpose of art, the other question is whether a photorealistic painting is actually more "lifelike?"

Sure, it represents the world exactly the way that a camera sees it, but does it represent the world exactly as we see it? With our human brains we are filling in details, we are focused on the subject of the painting, we are distorting, we are re-mapping shades of color to what we know they ought to be. "Lens blur" in particular isn't remotely the way our eyes perceive the world: our eyes can never roam around the unfocused parts of our own image, rather we ignore the blurry parts unless we specifically try to notice them, and anything we try to focus on becomes sharp.

So I appreciate the authors recognition that Piranesi's perspective may indeed be closer to how we perceive the world than "correct" perspective.

[bonoboTP]

> rather we ignore the blurry parts unless we specifically try to notice them, and anything we try to focus on becomes sharp.

Unrelated to perspective: Artists often put more detail in places that you're "supposed to" look at, and leave other things more as a blurry sketch, and our eyes naturally don't linger on those parts. What's interesting to me is that I can often recognize AI art by the way it doesn't make use of this, it tends to make everything equally detailed.

[jcalx]

I learned this back when I used to make maps for Source games (Team Fortress 2, CS:Source, etc.) — at first I sprinkled props and detailed geometry equally throughout the level, but it's better (in terms of art direction, player attention, and performance optimization) to put details in places that "matter". Valve does this [0] in their official maps and it's actually hard to notice until you really pay attention to it.

[0] https://nodraw.net/2010/08/tf2-density-of-detailing/

[dhosek]

One of the most dramatic things in this respect is to look at a photograph of a flame. Our lived experience of what flames look like is very different. The same is true of nearly any dynamic object in life (flowing water is another great example where the photograph doesn’t look like what we experience).

[semi-extrinsic]

Those examples are very challenging because they are volumetric objects, in addition to being dynamic. The flame does not even have a well-defined surface. What you see as red/yellow in the flame is in fact just glowing soot particles, and the point where this disappears is just a temperature difference in the gas such that the soot stops glowing.

[roughly]

One of my favorite books is "Seeing Like a State"[1], which talks about how the modern world has transitioned from talking about things like distance and space in human terms ("an hour's walk", "six bushels' worth of crops") to a surveyor's terms ("3 miles", "5 acres"). In the modern world, we're very used to thinking about things as mediated by technology (maps, cameras, compasses), and that affects how we think about arts, as well - matching what a camera produces is "photorealistic", but isn't how we actually _see_ things. This is one example, another I've always found fascinating is that the large Chinese landscape paintings play a trick where the perspective shifts depending on where you're looking - the whole work does not share a single perspective point, but any given point on the scroll looks "correct" for that spot and its surroundings.

[1]https://bookshop.org/p/books/seeing-like-a-state-how-certain...

[kqr]

I strongly agree.

> The distinctive feature of Piranesi’s perspective trick is that when you have a series of similar objects receding into the distance, such as houses or arches in a bridge, the nearer versions are just drawn as larger versions of those in the distance

This makes so much sense. When nature presents me with a series of identical arches, that is how I'm going to interpret them. Not as the geometric shapes that are actually projected onto the backs of my eyes.

[Etherlord87]

You make good points, but if we really perceived reality differently than cameras, someone would write an article about that - after all you can take a photo on your smartphone and compare to your IRL view. Yes it's hard to make it match but it's possible. Since no one made an experiment to show this, and since we understand pretty well the optics of a human eye, I think the cameras actually reproduce human vision decently. The aspect of a psychological bias and filling stuff in perhaps distorts in each possible way, so the camera shows the average human perception.

[WantonQuantum]

One thing that cameras can do that we can't do with our (unaided) eyes is change focal length. When a camera zooms in or zooms out, there is a change in the geometry of perspective. When zooming in, the change is difficult to see but is well documented. Portrait photographers are well aware of this because it tends to make people look better. When zooming out, especially when zooming out to a very wide field of view, the effect can be very noticeable. When people use a phone camera to take a picture of a group of people, the people at the edges of the photo can look very flattened and unnatural. Again, this effect is well known and very much discussed and written about in photography and art. In the article about Piranesi's work, we see a very interesting remedy to the perceived distortion of a wide view. Again, these works show a field of view, and therefore a focal length, that is not possible with our eyes. We can visualise in our mind's eye what the scene looks like but what our eyes are doing is moving around the scene while our visual cortex and ultimately our mind's eye construct a mental model. In our minds, we know that the arches of the bridge are the same size and shape despite being distorted by distance. We can look at what's in front of us and see that in the receding row of arches the nearest appears to be bigger than the farthest but we also know that's just an artefact of our point of view. Things are subtly different when we look with our eyes at a photograph, painting or drawing. It's not the same as when we see something in real life, we're looking at a 2 dimensional representation of a 3 dimensional scene. We do see the perspective in the picture - we don't see a bunch of lines on a two dimensional plane. Our brains are making that happen and are making some adjustments along the way. It just so happens that something we have difficulty adjusting so it "looks right" is a very wide angle scene. That's because we don't ever see a very wide angle scene in one glance in real life.

[kqr]

> When a camera zooms in or zooms out, there is a change in the geometry of perspective.

This is the opposite of what happens. The perspective is exactly the same, only cropped to a narrower field of view.

Changed perspective would be counteracting the lens zoom with foot zoom -- now this causes the change in perspective you talk about. But it's caused entirely by moving the camera. The zoom is just to preserve framing and could just as well be accomplished by cropping in post, assuming sufficient resolution.

[WantonQuantum]

You are 100% correct! I was unconsciously assuming the foot zooming. Thanks for picking that up!

[nyanpasu64]

I think foot zoom is the primary reason for faces looking weird (big noses) in smartphone selfies, while "a photograph occupying less of your FOV than the camera was capturing" is why people at the edge of a photo look distorted.

[bazoom42]

Just one well known difference: Human have two eyes and we use the difference to determine distances. A photo is flat.

Another difference: In human sigth you can only focus on the center of the vision field, while on a photograph you can look at unfocused points and observe the distortion on the edges of the picture. Human observation (like looking at the bridge in the pic) is really a series of observations where we move the focus.

[bazoom42]

Notably, photorealism is called photorealism because it deliberatley tries to reproduce photography, not because it tries to reproduce observable reality.