What is the best lens to take an accurate photo of the face with no distortion?

B

bossman101

Messages
4
Edit My Images
Yes
I measure facial proportions and ratios for cosmetic procedures. i’ve noticed that using photos for this can be difficult as measurements can be extremely off due to lens distortion. What would be the best way of capturing the face, as it is, so that the measurements would be the same as measuring from the physical face itself?

A photography enthusiast friend of mine mentioned the Canon 85mm F1.2 lens, while others on the internet have said that the rear camera of a good phone will suffice, if the photo is taken from far enough of a distance. Could anyone confirm whether either statement is correct, and if not provide other recommendations to get the most accurate image possible?

Many thanks
 
Hi and welcome to TP

I will await those who do portraiture to chip in but.......

Would you care to define "cosmetic procedures"?

As in, are you talking about make-up artistry or cosmetic surgical procedures?

My perception in the former is that good portraiture (choice of lens & lighting) will yield sufficient 'accuracy' but if the latter, again my perception, is that you are entering the realms of clinical/medical photography & illustration where knowledge of anatomy (as well like above choice lens & lighting to reveal facial structure) is an element in photographically recording the facial structure for the surgical procedures planning.
 
Type "focal length facial distortion" into Google and then click on images.
 
Box Brownie said:
Hi and welcome to TP

I will await those who do portraiture to chip in but.......

Would you care to define "cosmetic procedures"?

As in, are you talking about make-up artistry or cosmetic surgical procedures?

My perception in the former is that good portraiture (choice of lens & lighting) will yield sufficient 'accuracy' but if the latter, again my perception, is that you are entering the realms of clinical/medical photography & illustration where knowledge of anatomy (as well like above choice lens & lighting to reveal facial structure) is an element in photographically recording the facial structure for the surgical procedures planning.
Click to expand...
Thank you! Procedures are can be surgical, i.e. facial implants but also non-surgical, i.e. fillers. 3D scans of the facial bones etc. are used for planning in surgical cases, of course. However, we want to analyse facial ratios and proportions using photos for a better informed surgical plan using front and side profile photos.

There are quite a few people who analyse these ratios and proportions as a service who insist that taking a photo from a phone camera would suffice, but it would make sense for them to say that so as to not put off potential clients with the hassle of hiring a professional photographer. I am hoping to find an objective answer to gain the most precise measurements. I've heard different things about different focal lengths and apertures so just trying to see what the consensus is here.
 
bossman101 said:
Thank you! Procedures are can be surgical, i.e. facial implants but also non-surgical, i.e. fillers. 3D scans of the facial bones etc. are used for planning in surgical cases, of course. However, we want to analyse facial ratios and proportions using photos for a better informed surgical plan using front and side profile photos.

There are quite a few people who analyse these ratios and proportions as a service who insist that taking a photo from a phone camera would suffice, but it would make sense for them to say that so as to not put off potential clients with the hassle of hiring a professional photographer. I am hoping to find an objective answer to gain the most precise measurements. I've heard different things about different focal lengths and apertures so just trying to see what the consensus is here.
Click to expand...
Hmmm!
IMO you are looking at a specialist area of photography and as such, on the surmise you are in the UK here are two organisations (both IIRC have some international reach) where you might be able to 'talk' with specialist photographers......none too sure if I recall any TP members that have such specific experience.

www.imi.org.uk

Institute of Medical Illustrators - Setting & maintaining standards since 1968

A personal knowledge hub for medical creative professionals. Settings standards for the disciplines of medical illustrators within healthcare.
www.imi.org.uk www.imi.org.uk

rps.org

The Royal Photograpic Society Medical Group

rps.org rps.org

PS on a side note, if you or the service organisation you work for will be the supplying such imagery analysis, I wonder what is the "liability aspect" in regard to medical negligence claims based on incorrect or misleading analysis. No doubt there is insurance for such liabilities.
 
Generally speaking a focal length between 70mm and 100mm on a "Full Frame" camera is considered to produce a natural / flattering image.
However, this is based on viewer perception rather than millimetric precision for measurement purposes.
If you are wanting to collect data to inform a surgical procedure then I don't think photography alone is the answer.

You quoted an 85mm F1.2 lens...a beautiful portrait lens certainly, but you won't be needing the F1.2 bit so you could save yourself £1,000 or so there.
 
Last edited:
50mm is the closest to a human eye's view therefore I'd go for that on full frame (or 35mm on crop or 25mm on M4/3)
 
Its not the lens that’s giving rise to the apparent distortion, it’s the viewing distance – perspective comes from viewpoint.

If you choose to use a wide angle lens, you need to be close to the subject o fill the frame. This results in the “distortion”

Its generally accepted that an 85mm, or so, lens give a pleasing portrait. However, for your purposes, you may want to consider longer – say 200mm – to get further away.



You wont get a geometrically accurate result until you take the shot from infinity – not a practical proposition!
 
For this sort of thing, wouldn't you want, at least, three images: left, front and right, all taken from the same distance?
 
AndrewFlannigan said:
For this sort of thing, wouldn't you want, at least, three images: left, front and right, all taken from the same distance?
Click to expand...
Yes, that's right.

yzfmike said:
Its not the lens that’s giving rise to the apparent distortion, it’s the viewing distance – perspective comes from viewpoint.

If you choose to use a wide angle lens, you need to be close to the subject o fill the frame. This results in the “distortion”

Its generally accepted that an 85mm, or so, lens give a pleasing portrait. However, for your purposes, you may want to consider longer – say 200mm – to get further away.



You wont get a geometrically accurate result until you take the shot from infinity – not a practical proposition!
Click to expand...
Graham W said:
Generally speaking a focal length between 70mm and 100mm on a "Full Frame" camera is considered to produce a natural / flattering image.
However, this is based on viewer perception rather than millimetric precision for measurement purposes.
If you are wanting to collect data to inform a surgical procedure then I don't think photography alone is the answer.

You quoted an 85mm F1.2 lens...a beautiful portrait lens certainly, but you won't be needing the F1.2 bit so you could save yourself £1,000 or so there.
Click to expand...
After asking numerous people, I keep getting the same two responses:

1. The higher the focal length, the more precise the photo will be. Others say focal length isn't that important, but distance is, and so a higher focal length lens will allow us to get a suitable distance (something like 5 metres away, for example)
2. A focal length between 70-100mm produces a natural image of the face. Going too low or two high can cause different types of distortion, as the image below suggests

1701989172457.png

Opinions are roughly split evenly so I'm not sure who's right :D. Please advise!
 
For taking measurements I would think photogrammetry would be the only suitable option. It's more involved but maybe not as much as you might imagine once you have a workload.
One of the Welsh universities did some great work with artifacts from the Mary Rose (including human remains like skulls). These are used by academic institutions around the world to take forensic measurements from.
Sadly I forget the guys name that headed up the project or even which university he was from. I know he addresseed a conference to describe the work and process as it was the best photogrammetry at the time, maybe Severn years ago.
I hope you can find him from that description, if I find more info I'll update this.

In fact here it is. Maybe reach out to Nick Owen at Swansea
 
Last edited:
"so that the measurements would be the same as measuring from the physical face itself?" I think this is the critical bit. I doubt conventional photography will provide the results you're hopeing for. I would be looking at some sort of 3D scan.
 
bossman101 said:
I measure facial proportions and ratios for cosmetic procedures. i’ve noticed that using photos for this can be difficult as measurements can be extremely off due to lens distortion. What would be the best way of capturing the face, as it is, so that the measurements would be the same as measuring from the physical face itself?

A photography enthusiast friend of mine mentioned the Canon 85mm F1.2 lens, while others on the internet have said that the rear camera of a good phone will suffice, if the photo is taken from far enough of a distance. Could anyone confirm whether either statement is correct, and if not provide other recommendations to get the most accurate image possible?

Many thanks
Click to expand...
As others have suggested, you really need some specialist help with this.

A lot depends on how accurate the results need to be and the "exact" expectations of the user. Whatever you do, unless you look at photogrammetry techniques, and even then this comes with limitations, you are stuck with the issue of turning a 3d object into a 2d image. Even flat surfaces have their problems when trying to make measurements from photographs.

So any measurements on areas of the face in the photograph that aren't from a naturally flat part of the face isn't going to be accurate, But, with a little thought about working distance, lens focal length and care with how different software might be "correcting" distortion with different lenses might still give results that are "fit for purpose".

But this can't be a unique problem, and the search that DemiLion suggested threw up this reference, which seems a good starting point for further research.

 
bossman101 said:
I measure facial proportions and ratios for cosmetic procedures. i’ve noticed that using photos for this can be difficult as measurements can be extremely off due to lens distortion. What would be the best way of capturing the face, as it is, so that the measurements would be the same as measuring from the physical face itself?

A photography enthusiast friend of mine mentioned the Canon 85mm F1.2 lens, while others on the internet have said that the rear camera of a good phone will suffice, if the photo is taken from far enough of a distance. Could anyone confirm whether either statement is correct, and if not provide other recommendations to get the most accurate image possible?

Many thanks
Click to expand...
To answer your first question, the Canon 85mm f/1.2 is one of the best lenses money can buy in the DSLR world for artistic photography, but the f/1.2 aperture is utterly useless in your case. You don't want a creamy bokeh, you want accuracy.

With respect to smartphones, they are wonderful tools for candid shots, family souvenirs, etc., but they are not at all designed to be used in studio-like, scientific endeavours like yours. Previous answers gave you plenty of good advice. I'd just add that, whatever camera and lens you use, you'll need a proper studio flash unit and probably a tripod. You couldn't be further removed from the world of smarphones.
 
You have TWO main considertions...
1. INDUCED PERSPECTIVE DISTORTION...this is caused by CAMERA POSITION being too close (and NOT by the FL!!!). Examples this can be seen in comparing pohtographs of the same person, using the wrong FL lens. Usually a shooting distance of about 10' yields the perspective that we are accustomed to seeing a person's face. For a head & shoulders portrait that camera position comes with use with about 100mm FL on FF camera. The Steven Eastwood Photography web site had a comparison the facial perspective that comes with using sorter FL and longer FL.
Eastwood.jpg

Unfortunately that website no longer exists, but fortunately I had grabbed a screenshot prior to the website elimination.

2. OPTICAL DISTORTION...concept illlustrated in Post 10 of this thread. This is dependent upon the design of the lens...it is not FL specific, but usually comes at the shorter end of the FL range in a zoom lens.
 
Last edited:
Scenes of crime standard used to use a 100mm on full frame, and a 150mm on a 6x7. They would include a scale so prints could be matched life size (I've done some SOC)
 
A few years ago, I attempted to overlay a "modern" picture of a bike, taken with a crop DSLR on top of a scanned copy of an old 35mm negative of a bike taken with a 50mm lens - this was to compare frame geometry (lengths and angles). The old photo appeared to be straight on from the side. I couldn't manage it - despite distances, focal lengths and shooting position, I couldn't get the perspective angle quite right - if you looked at the shots side by side, they looked fit for purpose but attempting to overlay illustrated the problems - for one, it was almost impossible to get the wheels perfectly round or matching in distortion.

I reckon that even with the best suited lens, format and distance, you're going to struggle to get consistency between shots (beyond a single session with a fixed chair and tripod) to make comparative measurements. I suppose a lot depends on how precise and repeatable you want the results. I reckon that even holding a ruler in front of/next to the face would give different results.
 
weybourne said:
I reckon that even with the best suited lens, format and distance, you're going to struggle to get consistency between shots (beyond a single session with a fixed chair and tripod) to make comparative measurements.
Click to expand...
I imagine that one way to achieve consistency would involve something like a wire frame attached to the camera by a measured pole.

You would then place the wire at the same distance from a chosen feature (tip of nose, tips of ears) for each subject. This would have a similar effect to the chin rest used by opticians for some types of eye test.

It depends exactly how consistant you need/want to be.
 
Having worked in the area of biometric identity capture and verification, I must agree with a point made above that you need to consider a 3D view and take into account relative distance between the camera and the sitter of the different elements of a "face". Phone cameras are perfectly adequate for this if loaded with an appropriate biometric analysis app, which regular photography apps are definitely not. There are various companies offering biometric analysis apps for use on phones, I'm not able to recommend any particular one, but I will say having tested them for Home Office and Ministry of Justice purposes, there are about 6 companies that can do it well, 4 of whom are British.
However if you are prepared to accept the risk of getting it wrong using regular photographic kit, the standards described by @swanseamale47 above would seem to be the best yardstick.
 
weybourne said:
A few years ago, I attempted to overlay a "modern" picture of a bike, taken with a crop DSLR on top of a scanned copy of an old 35mm negative of a bike taken with a 50mm lens - this was to compare frame geometry (lengths and angles). The old photo appeared to be straight on from the side. I couldn't manage it - despite distances, focal lengths and shooting position, I couldn't get the perspective angle quite right - if you looked at the shots side by side, they looked fit for purpose but attempting to overlay illustrated the problems - for one, it was almost impossible to get the wheels perfectly round or matching in distortion.

I reckon that even with the best suited lens, format and distance, you're going to struggle to get consistency between shots (beyond a single session with a fixed chair and tripod) to make comparative measurements. I suppose a lot depends on how precise and repeatable you want the results. I reckon that even holding a ruler in front of/next to the face would give different results.
Click to expand...
I appreciate what you’re saying but this should have been possible. Frustrating but possible.

To do it I’d have loaded the image into ps, and then tethered the camera to achieve a perfectly overlaid image.
 
my wife works with a dentist who does all sorts of reconstructive stuff and he uses about 80mm on a Tamron with a ring flash pretty sure
 
I came across this possibly informative article

www.danvojtech.cz

Amazing how focal length affect shape of the face

Its always difficult to answer what is the best focal length for a portrait. Here is small preview how different focal lengths can change look of a face.
www.danvojtech.cz

Oh, IMO (uninformed?) in common with any subject where measurements are involved you must have reference points in regard to distances and angles of view.....without known reference points included in the image, any synthetic (digital) measurements (AI based for example) are at best approximations and at worst pure guess work ;(
 
The short answer to the first sentence is ‘you can’t’.

There’s loads of longer answers above.

But bluntly; if a person is sat in front of you the easiest and most accurate way of taking measurements is to simply get a bloody ruler out. To do it with a photo for an accurate record requires the camera and a ruler.

If you use a fixed position camera and subject, take reference measurements, then you’d have something repeatable, but any imperfections in the following setup throws it out. You’re straight into having a margin of error.

Why bother?

In statistics we do sampling, which means we can extrapolate a narrative from a representative sample, and it’s pretty accurate, but it’s still not as accurate as measuring the entire population. In a world where you can count the entire population why rely on a sample and accept a margin of error.
 
Last edited:
@bossman101 I don't think it's possible hence the differing opinions. For accuracy, 3D models extracted from videos may be better.
The dentists I know use 100mm, and this focal length is advised by Canon in their guide for dentistry photography.
 
Thank you all for your replies. I must accept that it was rather naive of me to think that it would be easy to get an extremely accurate photo of the face using standard photography.

Facial ratios and proportions tend to be measured from the front profile or side profile alone, and so are all 2D measurements, and so I thought it would be appropriate to get as accurate a photo as possible. To do this, an 85-105mm lens at 2+ metres away would be best, but not perfect.


ilikebowens said:
For taking measurements I would think photogrammetry would be the only suitable option. It's more involved but maybe not as much as you might imagine once you have a workload.
One of the Welsh universities did some great work with artifacts from the Mary Rose (including human remains like skulls). These are used by academic institutions around the world to take forensic measurements from.
Sadly I forget the guys name that headed up the project or even which university he was from. I know he addresseed a conference to describe the work and process as it was the best photogrammetry at the time, maybe Severn years ago.
I hope you can find him from that description, if I find more info I'll update this.

In fact here it is. Maybe reach out to Nick Owen at Swansea
Click to expand...
Thank you for mentioning photogrammetry. I am actual aware of a medical imaging company that does something similar: https://www.cavendishimaging.com/3d-photography-clinical-measurements/ , but based on their example images, it doesn't seem very clear and it does make a few mistakes when producing the image. Nonetheless, I think for absolute precision, it is best to take individual shots of the front and side profile from a 3D image. Most patients do get a full head CT scan which does include the facial soft tissue overlayed on the bone, but I think a real 3D image will be superior, just for the sake of visually analysing it. Not sure if I'm trying to justify something unnecessary here though.

lindsay said:
Having worked in the area of biometric identity capture and verification, I must agree with a point made above that you need to consider a 3D view and take into account relative distance between the camera and the sitter of the different elements of a "face". Phone cameras are perfectly adequate for this if loaded with an appropriate biometric analysis app, which regular photography apps are definitely not. There are various companies offering biometric analysis apps for use on phones, I'm not able to recommend any particular one, but I will say having tested them for Home Office and Ministry of Justice purposes, there are about 6 companies that can do it well, 4 of whom are British.
However if you are prepared to accept the risk of getting it wrong using regular photographic kit, the standards described by @swanseamale47 above would seem to be the best yardstick.
Click to expand...

Thank you for your contribution too. I'm not entirely sure what you mean by "biometric analysis app", as there are of course facial recognition apps but they would not provide me with an actual image of the face? I've tried looking and I'm not entirely sure.

There are of course apps that can 3D scan objects. I haven't tested them out yet, but I have an iPhone 14 Pro so maybe it will suffice. I think I'll try and scan my own face and figure out if the measurements are to scale. If not, I'll have to opt for professional 3D scanning, or a 100mm lens
 
Thanks for responding in the thread @bossman101 .
My point about biometric apps relates to facial recognition technology, which you are quite right in saying does not actually capture an image of the face - instead it takes lots of vectored measurements to create a profile based on the relative positions of facial features and landscape. I was thinking that this is what you needed, but maybe I was overthinking the task.
 
lindsay said:
... instead it takes lots of vectored measurements to create a profile based on the relative positions of facial features and landscape.
Click to expand...
As a slight aside, we watched one of Hannah Fry's programmes last night.

This one was mainly on the subject of facial recognition and it included a long subsection on the early efforts at facial recognition, which were sponsored by the C.I.A.. I hadn't realised before, how little data is required for a reliable match, when using basic physiological measurements.
 
...which is why we normally use other things like "liveness" checks to verify that what the camera sees is a real face. We have tested using photographs, photos held in front of a real face with eyeholes cut, videos of faces, deepfakes using silicon masks and even silicon modelled heads, in testing and partnered with similar techniques to match fingerprints to faces, etc etc. This to verify identity of course. It all depends on the application for which it is being used though - my role testing this was related to passports, immigration enforcement and the evidential level needed for criminal charging.
 
bossman101 said:
I measure facial proportions and ratios for cosmetic procedures. i’ve noticed that using photos for this can be difficult as measurements can be extremely off due to lens distortion. What would be the best way of capturing the face, as it is, so that the measurements would be the same as measuring from the physical face itself?

A photography enthusiast friend of mine mentioned the Canon 85mm F1.2 lens, while others on the internet have said that the rear camera of a good phone will suffice, if the photo is taken from far enough of a distance. Could anyone confirm whether either statement is correct, and if not provide other recommendations to get the most accurate image possible?

Many thanks
Click to expand...
I use a Nikon 85mm with my full frame D750.
 
Gotta say this is one of the better 1st posts by new members we've had recently. Welcome bossmann.

I would chime in and say there is no abolutely right answer, as my thoughts were along the lines of using a lens that gives the same perspective as the human eye. On 35mm this is 43mm, but as has already been pointed out to achieve the results you are after you would have to go right into your subject and the image becomes somewhat distorted. just the same as if you would invade someone's personal space. A lens in the 85-105 range at the sweet spot aperture wise may be the best option....
 
welshwizard645 said:
along the lines of using a lens that gives the same perspective as the human eye. On 35mm this is 43mm, but as has already been pointed out to achieve the results you are after you would have to go right into your subject and the image becomes somewhat distorted. just the same as if you would invade someone's personal space. A lens in the 85-105 range at the sweet spot aperture wise may be the best option....
Click to expand...
The difficulty of the expression 'same...as the human eye' is that 'perspective' -- the relationship of main subject to objects relative to subject position' -- are determined not at all by FL, but determined simply by the camera POSITION. Similarly the 'same perspective as human eye' is determined entirely by where the person is standing at a particular time! I took the same photo from one camera position using three different FL lenses, and cropped to the identical area during post processing, and the three crops had identical 'perspective'...proving it was NOT FL than provided the 'perspective'.

This was extracted from a post on another forum, many years ago
Below, I 'enlarged' the three shots during post processing, so that all the items (like trees) seen in the frame are similarly sized in all the photos. The point of this series is to illustrate that Focal Length has NOTHING to do with 'perspective', as perspective is only changed with camera position. In effect, the framing is affected with a change of FL without change of camera position, which is not changed for these three shots. Framing differences are neutralized via cropping the original shots (only the shot made with 200mm FL actually shows entire frame).​
​
IMG_200mm-3b.jpg
​
​
IMG_97mm-2b.jpg
​
IMG_55mm-1b.jpg
​
In effect, the framing is affected by a change of FL without change of camera position (framing difference neutralized via these crops of the original 97mm and 55mm shots)​
​

Some past things posted on the web have claimed 'like a 50mm lens', but provide little substantiation for that specific FL value.
Some things posted on the web have tried to equate the eye's FOV, but the eye has good vision within only a narrow central cone, and ability to see words clearly to the right/left of the one eye that one fixes their vision to is very poor indeed! And the total FOV of the eye includes a wide swath of recogition only because of our familiarity with out surroundings, experienced over time, but I could not tell you speicificall by name what knickknacks my wife placed to the sides of our TV 3-5' to the side, if staring at the TV screen!

Just as the above three photos demonstrated 'same perspective' regardless of FL, the issue with trying to assess identitiy by relative positions is the face is 'distorted' by camera position (as illustrated in post 18) and one set of measurements taken from a flat 2D rendition of the face is only accurate at ONE camera position...FL used only alters how loose or tight we are framed on the face.
 
Last edited:
You must log in or register to reply here.
Back
Top