6

I know in a portrait photo we use distance between camera and the subject in such a way as to create a pleasing and hopefully natural-looking perspective. I think it's good on 1 ~ 1.5 meter to get a best portrait photo.

I'm computer engineer and have a constraint in distance. The person (subject) and camera should be at 50 cm distance.

Is there any camera or lens or any hardware filter that I can use to get a best portrait (head-shot) and prevent perspective distortion with this distance limitation?

Update 1: this is my figure captured but I think all of them is not good as well (perspective distorted):

headshot 1 headshot 2 headshot 3

xiota
  • 26,951
  • 4
  • 39
  • 126
Akj
  • 79
  • 1
  • 6
  • 4
    What is your real problem? What are you trying to achieve? Is it some sort of computer-vision applicaiton, or photo booth, or what? In other words, why is it important that "I'm computer engineer and have a constraint in distance"? Do you want to take self-portrait using your laptop? – aaaaa says reinstate Monica Jun 11 '18 at 22:09
  • 2
    Constraint is a selfie with distance restricted to arm length   No lens can correct for perspective because perspective is strictly a function of distance (as in answers by @mattdm @WayneF). But it might be doable with mutiple lenses and some 3D modeling. (Unaware of any such existing software.) – xiota Jun 11 '18 at 22:35
  • Yeah, I think even if this is inspired by a computer vision problem, the selfie case is a pretty broad one with general application. (And it's actually one of the things specifically considered in the paper linked in my answer.) – mattdm Jun 11 '18 at 23:00
  • 1
    @mattdm really this is a computer capture program that need portrait picture but real and nice one and store image for rhinoplasty use. but this is a photography question not computer engineering question. Is it clear? help me to improve my question. – Akj Jun 12 '18 at 08:17
  • 2
    @Akj You need to consider your objectives. If this is a technical photograph that is being used for a technical purpose then don't worry about aesthetics and concentrate on capturing the technical information that is required for the application. If you also want to capture an aesthetically pleasing portrait of a rhinoplasty patient after the procedure, simply for marketing or patient feedback purposes, then just take a second photograph with the correct equipment for that purpose. – J... Jun 12 '18 at 14:19
  • @J... you are right. this is a nice question, nice place and nice answer. this is not a general question or simple question. consider a portrait photo for rhinoplasty that near object like as nose is in deformation or preceptive distortion. you should do in ~ 50 cm. this is the main problem. I think there is no device in the world can do this. – Akj Jun 12 '18 at 15:40
  • Please add an explanation of why you have this issue in the first place. – JonathanReez Jun 12 '18 at 19:28
  • @JonathanReez I add the pictures, what should I add more? – Akj Jun 12 '18 at 20:47
  • 1
    No, as in explain why you can't have the camera farther away. – JonathanReez Jun 12 '18 at 20:48
  • @JonathanReez because there is a room, and the space is not allowed for this. this is same for 10 rooms that need this capabilities. – Akj Jun 12 '18 at 20:53
  • Personally, I don't think we need three separate examples. One would make the point just fine. Pick your favorite. :) – mattdm Jun 12 '18 at 22:19
  • 1
    @Akj but what is your objective? To take an aesthetically pleasing portrait of a person (ie: just a picture that is nice to look at?), or to take a technical photograph that captures dimensional information about the geometry of their face? (something that will be used for the medical procedure?). These are completely different objectives. – J... Jun 13 '18 at 00:24
  • Just a good portrait photo for rhinoplasty use. @J... – Akj Jun 13 '18 at 08:21
  • @Akj A good portrait for whom? Do you understand what I am asking? – J... Jun 13 '18 at 09:43
  • @J... a good portrait for doctors. nose and other object be like as ~1 m distance. – Akj Jun 13 '18 at 11:56
  • @Akj Is perspective distortion really that important, then? If it is just for doctors to make an assesment they should be able to get the information they need. Perhaps the best option here would be to take three photos - direct, 45-degree, and side profile. This would overall provide more and better information. You could use a simple lens correction algorithm to remove the worst aspects of the lens distortion - this won't correct perspective but it will produce a reasonably "flat" image. – J... Jun 13 '18 at 12:28
  • @J... direct means direct, 45-degree is ok, but what about side profile? is there any possibilities shows a diagram for example via paint? – Akj Jun 13 '18 at 12:54
  • @Akj 90 degrees to camera. This way the nose is in the image plane so it won't suffer from lens effects, plus you get a better diagnostic view for the surgeon. – J... Jun 13 '18 at 23:40

2 Answers2

25

There is no hardware solution, short of some crazy* arrangement of mirrors to extend the actual optical distance. That's because perspective distortion solely related to distance, but there might be a software one, if you're able to throw a computer at the problem and able to accept some limitations.

For a computational approach, see this paper: Perspective-aware Manipulation of Portrait Photos (pdf link) — the basic idea is that you build a 3D model of the head and then use that to distort the image to match a photograph taken from a different distance. Unlike many computer graphics or computational photography papers, this one is quite readable by the layperson (e.g. me), although I admit to glossing over the actual hard part about the tensor model.

@Ian points out that there is an online demo. I ran your example through that, with this result:

animation of the magic

This solution uses a single image; I'm not sure if that's part of your constraint. Having more information for doing this kind of sorcery in software is one of the uses for multiple simultaneous cameras — think some of the new smartphones.

* If you're got a fixed location, it doesn't actually have to be so crazy. Place a large mirror pointing up at a 45° angle against the wall. Mount your camera near the ceiling. Keep the mirror clean!

mattdm
  • 143,140
  • 52
  • 417
  • 741
  • wow that paper is some cool stuff, from perspective of science. – Horitsu Jun 12 '18 at 04:58
  • 1
    Very good article! There is a link to an interactive website where you can actually try out this program with demo photos or your own: http://faces.cs.princeton.edu – Ian Jun 12 '18 at 06:39
  • @Horitsu I see what you did there :) – mattdm Jun 12 '18 at 20:32
  • @Ian I see your demo, you means this calculate by software or anything else? – Akj Jun 12 '18 at 21:41
  • @Akj The software is a demo by the authors of the paper. It is a software approach. – mattdm Jun 12 '18 at 21:51
  • @mattdm yeah I see. anyway thanks. this is a good one but maybe hardware based is better. – Akj Jun 12 '18 at 21:52
  • @Akj Well, except, note what I said at the beginning: there is no hardware solution. You might be able to add a little distance with a mirror. – mattdm Jun 12 '18 at 22:29
  • For that mirror solution: Attention! You have to use a mirror with the reflecting surface on the front not on the back! With reflecting surface on the back you will get double reflections from the glas and the reflective surface. So you will need a (often expensive) mirror with reflective side on the front, that is very sensitive and maybe also not easy to clean. – Horitsu Jun 13 '18 at 05:24
  • @Horitsu you are right. I think mirror is not works, am I right? – Akj Jun 13 '18 at 11:15
  • @Akj such front side mirrors can absolutely work. on my old university they used such mirrors so project with beamers onto big walls while the room was too small to do this in a direct way. even with polarized light. – Horitsu Jun 13 '18 at 11:19
  • @Horitsu is there any picture or setup for this mirror? – Akj Jun 13 '18 at 12:05
12

Classic standards have always been more like 2 meters Minimum distance for portraits. 3 meters is common in professional studios, for even better perspective. Too little distance (like 1 meter) will exaggerate and enlarge closer features like noses. This distortion will not please the subjects.

The standard notions are the 105 mm lens is good for head and shoulders portraits on full frame 35 mm film, simply because the field of view will demand and force the proper minimum distance for good perspective. And for example, a crop factor 1.5 camera would use the equivalent 105/1.5 = 70 mm equivalent focal length for proper perspective. The distance will be the same either way.

Perspective is NOT about the lens. Perspective is ONLY about the distance, i.e., where the camera stands, and the view it necessarily sees from there. If you want to improve perspective, you must change where you stand. The chosen lens might change where you must stand (for the desired field of view), but the perspective result is only about where you stand.

So stand back some. Zoom in all you want for the view you want, but stand back some if you want proper portrait perspective. Any kind of proper planning will include standing at the necessary distance. Perspective is the view seen standing at that distance.

WayneF
  • 12,879
  • 1
  • 16
  • 30
  • you means there is no way, no technique, nothing else for this close distance? am I clear? – Akj Jun 11 '18 at 20:38
  • I doubt about the meaning of "The chosen lens might change where you must stand" so you means with specific lens we can stand near? – Akj Jun 11 '18 at 20:41
  • 3
    The chosen lens can affect the size of the field of view, i.e., maybe creating a head and shoulders view. But only the distance controls the perspective, which can avoid enlargement of the near features (like the nose) caused by standing too close. – WayneF Jun 11 '18 at 20:45
  • Aha, I get it... – Akj Jun 11 '18 at 20:46
  • 3
    To be fair, a "lens" could in principle create distance greater than any external dimension of the lens if it's actually a system of mirrors and lenses by bouncing light rays back and forth multiple times. – R.. GitHub STOP HELPING ICE Jun 11 '18 at 23:05
  • @R.. But only if the first refractive surface (or curved reflective surface), as opposed to flat reflective surfaces, is placed at the other end of all of those mirrors. – Michael C Jun 12 '18 at 07:55
  • 2
    @R Thus, the popularity of the "bathroom mirror" selfie. – Michael C Jun 12 '18 at 07:57
  • As long as you can crop enough you can even use a wide angle lens providing that it is corrected for the optical distortion, here is a comparison I did with a 14mm and a 500mm: https://ttbek.deviantart.com/art/Example-wide-angle-telephoto-perspective-659407670 The distance is indeed all important for the perspective distortion, though I have trouble convincing people of this on other photography forums as they are just so dogmatic in the generalized rules of which focal lengths to use for what thing that they usually insist focal length affects perspective distortion. – ttbek Jun 12 '18 at 09:49
  • @MichaelClark is there any such lens in real store? – Akj Jun 12 '18 at 12:35
  • I'd love to know the answer to that question too. Maybe it should be its own question. – R.. GitHub STOP HELPING ICE Jun 12 '18 at 20:10
  • Added as a new question: https://photo.stackexchange.com/q/99369/5007 – R.. GitHub STOP HELPING ICE Jun 14 '18 at 23:54