Perspective (yet again, sorry!)

Started 2 months ago | Discussions thread
Just Tim 4
Just Tim 4 Contributing Member • Posts: 621
Re: books are frequenty wrong

tony field wrote:

tko wrote:

Tom Axford wrote:

I have an old copy of "The Manual of Photography", 1978 Edition (Focal Press) and with all the discussion recently about perspective, I thought I would look at what it had to say on the matter.

Here are two extracts from the section on perspective (in Chapter 4 "The Geometry of Image Formation"):

Perspective on taking a photograph

As we have already seen, the perspective obtained on taking a photograph - sometimes termed the true perspective - is governed solely by the viewpoint. If the viewpoint is fixed, there can be no change in perspective, even if we change to a lens of different focal length - although, as we have seen earlier, the image size will alter. If, however, the viewpoint is altered, so will be the perspective, and no change of lens will recreate the perspective obtained at the first viewpoint.

This is correct, thanks


Perspective on viewing a photograph

The perspective obtained on viewing a print - sometimes termed the apparent perspective - depends, firstly, on the relative sizes of objects in the print - and hence on the perspective obtained in the negative on taking the photograph - and, secondly, on the distance at which the print is viewed.

Correct perspective is said to be obtained when a print is viewed in such a way that the apparent relation between objects as to their size, position, etc., is the same as in the original scene. This is achieved when the print is viewed at such a distance that it subtends at the eye the same angle as was subtended by the original scene at the lens. The eye will then be at the centre of perspective of the print, just as, at the moment of taking, the lens was at the centre of perspective of the scene.

Unfortunately, this is wrong

The book is correct. You are misinterpreting what it said. That is a superb and technically accurate book particularly so for the older editions. I wish I had a copy.

I would agree. The book is a technical manual and describes accurately how an image is mapped by a lens onto a 2D surface. What it doesn't describe in detail is how the human eye interprets that information and re-builds a 3D understanding. But it doesn't seek to to that and you must read it with that understanding. It is a common mistake to think that because something is *fixed* in an image that you see and understand it as *fixed*. In order to understand perspective fully you must understand the nature of the assumptions you make when you view images.

The viewing point for a photo is forever fixed at the point from where the photo was taken. In the 3D world, moving closer and farther changes the relative sizes of objects that aren't the same distance from the lens. But once captured in a photo, the relative sizes are forever fixed

Of course! That is the nature of projecting a 3D scene onto a 2d surface of the sensor

I agree, but add that just because the *relative* sizes are fixed it does not mean that your assumptions of scale and distance are.

Take the classic example of a portrait taken close up. The features are exaggerated. Try as hard as you can, no amount of moving the photo back and forth will alter this exaggeration. Printing the photo freezes the perspective, since there no "center of perspective" for a flat, perpendicular object

Well that has to be said with care.

This has to be taken with EXTREME care. Nothing is as familiar to the human eye as a human face, nothing is imprinted on our memories more clearly. Such is our pre-disposition to form a correct understanding of a human face that it's actually quite difficult to spot what should be glaringly obvious. And if you can't see the obvious clearly then do you really think that we see subtle changes in perspective clearly?

Just as a proof see if you can spot what's wrong with the image below:

Explain that this shot which is virtually full frame from about 1.5 and certainly less than 2 meter distance witha 28 mm lens. I do not see the that distortion you are describing

The perspective impression that you see on the 2D projection is a perceptual process and your perception might be surprisingly wrong in many cases because you misinterpret the visual clues.

Absolutely. In fact we ALWAYS misinterpret visual clues in images, we never see them correctly. But there is another process at work, *consistency scaling*. Basically we find it far easier to navigate in a world where perspective is *consistent* and so the brain works to maintain this consistency. It doesn't mean that we see a distorted view of the world or that our assumptions are inaccurate, just that they maintain consistent as we move through it. And they remain consistent from our position as observer. You can see this when you first get a pair of glasses. At first they make you a little dizzy, but you get used to them. Not many understand the full importance of this, that you brain recognises the change in focal length, (and the slight change in angle FOV), and how vanishing points appear to move at a different speed in your peripheral vision to your central vision and cancels the effect out. It learns how to adapt what you see so you have a consistent understanding where perspective stays constant.

This happens in images as well. We form a consistent understanding because our brains try to cancel out changes in perspective with distance. The illusion with the odd shaped room is a good example of this process. It only works, (and is designed to work), from one point, the centre of perspective. If we stand in every other place we form a remarkably consistent view of the shape of the room, it is only when we stand in the centre of perspective that another more likely interpretation is presented. The illusion of the room is designed to show that in cases where the most likely is obviously incorrect we still interpret perspective in terms of the most likely as dictated by our experience. And so we see the tow people as being of different sizes rather than the room being an odd shape, even when we know the reverse is true.

Images are the same, we form a consistent understanding of perspective until we reach the centre of perspective when another interpretation that is more consistent with our experience presents itself. It is not a linear process, and with all things we see you can't make the assumption that the maths is correct and that we always see this in absolute terms.

You can't view images without understanding the nature of the assumptions we make when we view them.

Take a photo. Now take a photo of the photo from far away and close up using a zoom the fill the frame. The two resulting photos will look identical.

Since you have move far away from your original shooting point of course the actual perspective changes in the final 2d image projection. The images are not looking identical.

Agree, they don't. But it is hard to detect this with the subject of a human face, especially one where there is no other perspective other than our assumptions of the shape of the human face.

Many classic works have serious flaws. In this case, the flaw is somewhat laughable. Suppose you take a photo of a mountain from a mile away. Do you have to view the image from a mile away to have correct perspective?

Of course not you would not have the correct ocular point of view if you tried to do that.

I agree, it's a case of *absolutes gone mad* - the relationship is to focal length x magnification not subject distance.

What about a macro? Do you have to glue the photo to your face?

Macro is a special case as we ar viewing the world in a way which we have not other experience of other than the photos we see. Besides it's related to the focal length x magnification (the focal length at the point of focus and not the number stamped on the barrel), somehow this has been twisted to mean the actual distance to the subject...

Okay, technically, the edges of a photo might be smaller as you move the image closer, since they are now farther away. But most lenses and the human eye correct for this effect, which can't be categorized as "perspective" effect.

Wrong again. Sorry.the correct ocular distance will always always give you a correct perspective effect.

I agree. And if people are still in doubt that they do not see or recognise the nature of the assumptions we make, and that those assumptions are relative to our position as observer, that because we see consistency relative to our position we make the assumption that it is the object that is consistent:

An image of simple perspective, a vanishing point that is in the centre of the image:

Now if the perspective is *fixed* in the image and is an absolute of the image then if we hung two identical copies on either side of a corner in a wall then wouldn't the *fixed* perspectives (or vanishing points) converge?

But they don't they remain resolutely fixed only to your position as observer, in fact they actually splay out in relation to your position:

BTW, turn the portrait the correct way up.

Post (hide subjects) Posted by
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow