Q and A ADVICE - FACE COMPARISONS
Anthropometry, Face Biometrics, Identifying Faces in Photographs
by Joelle Steele
QUESTIONS? Contact firstname.lastname@example.org
I try to answer all questions by Monday mornings at the latest. Not all questions make it into this column, but every reasonable question is answered by E-mail, so be sure your E-mail address is correct. You can use your browser's "find" or "search" function on the Edit menu to search this page.
Q & A Material as of AUGUST 22, 2017
Q. Would you please explain why you don't examine overlay-type comparisons.
A. Overlays are the most deceptive kinds of comparisons that can be made because people can look alike, and an overlay only accentuates their similarities. But to be sure they really are the same person, to accurately analyze and compare any two faces, each face must be thoroughly measured out and the facial proportions compared.
Q. Do you collect antique photographs of people? Are you interested in acquiring ones of famous people?
A. No, I don't collect photographs of any kind and I'm not interested in doing so.
Q. I see your name in posts where a collector says you don't know what you're doing. Why don't you respond to them? What are you afraid of?
A. I don't respond because I don't care about what collectors say. I've been doing this since 1980, and I've analyzed tens of thousands of faces for clients ranging from auction houses to law enforcement agencies. Collectors, on the other hand, don't make up even one percent of my business.
Q. How come you're not doing analyses of Billy the Kid, the James brothers, or Abraham Lincoln anymore?
A. Because I get so many of them every week, and they always come from amateur collectors who make up the minority of my clientele, yet somehow manage to waste the majority of my time attempting to argue with me about how I'm wrong and their photo really is a Lindoln or a Billy when there is not even the slightest resemblance and the measurements are not a match. I'm not a retired hobbyist with time on my hands to engage in these debates, so I have elected to not respond to these requests.
Q. Why did you say that my photo was not a match when face recognition software said it was? How can you argue with software that law enforcement agencies use?
A. Law enforcement agencies rely on face recognition software that uses a small number of facial points to search enormous databases containing millions of faces. Such a search could yield thousands of possible matches. To reduce the results, they filter the searches to include non-facial information, such as race, height, etc. This can bring up 10-100 possible matches. Unlike TV and movies where the computer finds an exact match, in reality someone has to examine all those possible matches manually (by eye). Me? I'm not searching a database. I already have the possible match because my client has supplied me with it, along with one to three photos of known faces of the person they think may be in their questioned photo. My only job is to measure, using ear patterns and up to 100 facial points, and then compare one face to the other three to determine whether or not they are an exact match.
Q. You emailed me within five minutes saying my faces were not a match. How can you tell so quickly whether someone is not the person they are purported to be?
A. I do a preliminary analysis (quick measurements) that rules out non-matches based on the proportion of the eyes to the face, the ear placement and patterns (when visible), the shape of the head and face, and the vertical alignment and proportions of the facial features, all of which are dependent on the underlying bone structure of the face and head. If any of these are not an exact match, then the two faces cannot be a match.
Q. What happens when you compare a face of a person when they are 35 with their face when they're 60?
A. The skin changes as a person ages, e.g., wrinkles, sagging, etc., and the ear lobes and nose tip may lengthen slightly. In the cases of antique photos, there may also be damage to the jawbones due to severe dental decay and tooth loss, which can alter the mouth and the overall face length. The skin changes are unimportant, so I just measure the bones as usual. But damage to the jawbone can impact on the measurements, so I have to pick other bony points to measure and compare, and sometimes this is not very reliable.
Q. I have two full-face, straightly aligned photos in which everything measures and matches perfectly except for the eyes, which have a slightly less than 4% difference in the eye width from endocanthion to endocanthion. Does that mean these faces are not a match?
A. If you are absolutely sure that the faces are not angled at all into a three-quarter view, then it is not a match. If you measured very, very carefully and double-checked your measurements and ratios, then this is not a match. A variance of more than one-half a percent is significant in a face measurement. Could the faces be those of siblings or other close relatives who resemble each other?
Q. I can't find a way to order the highly detailed report you used to offer. Do you still provide that service?
A. Sorry, I am scaling down many of my services, and I no longer offer that particular report. However, I still offer the quick analysis and quick analysis with brief explanation. That is all most people need or ask for, and all of the measurements are the same as those detailed in the big report.
Q. I sent you five exemplars and you still couldn't analyze my photo to see if it was a match. Why?
A. Your exemplars were fine, but your original was very faded and the features were indistinct, and I need a lot of details to be able to compare and confirm an identity of a face.
Q. Can you tell me whether I am related to someone by matching up similar features in our photographs?
A. No. People can look related and not be related at all, and people can be related and not look alike. Similarities can go from one generation to another or they can skip a generation (or two) entirely. The only way to know for sure if you are related to someone is through vital statistics documents and/or DNA testing.
Q. From your articles it looks like you never find a face match between photos. Why is that?
A. I actually find lots of matches for my clients, but my articles are designed to illustrate the problems of authenticating identities of faces in photographs, and I use famous people to do that. Unfortunately, there are very few photographs of those famous people, so I get a lot of contact from people who want to strike it rich with a previously unknown photo of a famous person, but their photos rarely match up.
Q. I loaned someone a copy of your 1992 book, Anthropometry and The Human Face in Photographs, and I'd like to buy a replacement. Where can I get a copy?
A. That book is no longer available. However, I have rewritten and expanded it into a much better book on the subject. It is titled Face To Face: Analysis and Comparison of Facial Features to Authenticate the Identities of People In Photographs.
Q. Why do you reject so many of my photos for analysis of facial features?
A. The single biggest obstacle to making accurate analyses and comparisons of faces in photographs is the quality of the photographs being analyzed. If I rejected your photos, it is probably because I cannot analyze and compare any photos that are 72dpi and less than 2" wide. That's why I ask for 600dpi scans so that I can enlarge the photos onscreen and see the details. The other reason I sometimes reject certain photos for comparison is because they lack sufficient details due to fading or overexposure.
Q. Can you compare paintings of two men the way you compare photographs? I recently inherited two miniatures and I think they are of the same person but I don't know how to correctly compare them myself.
A. Yes, I can and do compare painted portraits. The procedure for paintings is the same as for photos.
Q. Are the facial features analyses you provide to collectors the same as the ones you provide to appraisers?
A. All of my analyses are done in the exact same way, no matter who my clients are. However, I almost never work directly with collectors. The majority of my clients are appraisers, genealogists, private investigators, library/museum archives, and family historians – in that order.
Q. Are cephalometry, anthropometry, and biometrics the same as facial features analysis? What about physiognomy? Is what you do considered forensic or physical anthropology?
A. These are all variations on the same theme, but they are not exactly what I do. For example, anthropometry (biometrics) is the measurement of the face and body, including fingerprints and iris patterns. Cephalometry is a variation on anthropometry and involves the measurement of the face and head, usually from X-rays, and commonly used by dentists and plastic surgeons. Physiognomy, also called anthroposcopy, is not related to any of the above and is about the interpretation of facial characteristics to assess personality traits. In general, biometrics is the application of statistics to the measurements of the face and body. It is most commonly used today by law enforcement officials. I am an anthropometrist specializing in the analysis and comparison of facial features to authenticate or confirm the identity of a person in a photograph, so my practice is based on forensic and physical anthropology but is specialized and adapted only for use with photographs.
Q. Do you know how much skin damage, like pox marks, can be visible in photographs prior to about 1870?
A. With facial marks (scars, line, moles, etc.), they can appear as worse than they really were or they can be so faded that they don't show at all. This can be due to age of the person as well as the type of photographic process used, e.g., the lighting and the exposure time. In general, the greater the length of exposure, the more evident such marks will be. But don't forget make-up, since this was used by some individuals as well as photographers to mask facial flaws. These idiosyncratic marks are highly unreliable for comparing faces as they can appear suddenly in a person's life and they can just as easily get worse or disappear altogether. Bone structure proportions are the main thing to compare.
Q. How come you won't do a facial features comparison of my photos just because one of them was digitally restored?
A. Your image was not restored. It was retouched and enhanced. In restoration, the work is meticulously done by hand and all the tiniest lines and shading that were there originally are left intact. Your photo is so "clean" that it doesn't even look real. I need those original lines and shading for an accurate analysis because they are part and parcel to the facial features themselves. They indicate the depth, height, and breadth of a nose or chin or other feature. If you send me a 600 dpi scan of that photo — no adjustments at all — I will do your comparison.
Q. Why are all your examples for face analysis of famous people? What about ordinary people? Why don't you use examples of those?
A. Faces are faces, famous or not. I use famous faces when I teach face biometrics because they make my classes more entertaining. I try to do the same with my articles on this Web site.
Q. Can you identify people in surveillance photos?
A. Depends on how clear the surveillance photos are and whether you have good exemplars to which I can compare them.
Q. In an old article in a genealogy newsletter you mentioned using a software called FaceShell. Is that a program you created, or is it now called something else?
A. Wow, you are going waaaaaay back in time! FaceShell was an early form of face biometric software and it was definitely not designed by me. It could only be used for part of a facial features analysis, similar to most of today's biometrics programs, such as Visual Face Recognition software. The rest had to be done by hand. I don't know what happened to FaceShell. It was okay for ruling out the non-matches, but not adequate for authenticating an identity.
Q. Why do you charge for your facial features analysis services when there are plenty of people who will do it free?
A. I charge for my services because I put a value on my time and my expertise. If someone does it for free, it could be that you get what you pay for.
Q. Even though my photo says, "Mr. Lincoln" on the back of it, and it looks like Lincoln, you said it wasn't. How can you be so sure?
A. I know Lincoln's face very well and your image is not a match. Having a name written on a photo doesn't confirm the identity of the person in it. The man in your photo has a beard and is wearing a stovepipe hat, but that is where the resemblance ends. Someone probably wrote "Mr. Lincoln" on it because of that, probably as a joke. I have seen many photos labeled as people they are not for the very same reason, including numerous ones labeled "Bonnie and Clyde" and "Frank and Jesse James" that are definitely not the celebrated outlaws.
Q. In analyzing the angle of the eyes when comparing facial features of people in photographs, what exactly is being measured? How accurate is this as a comparison method?
A. The angle is measured at the corners of the eyes, which are called the medial and lateral canthi. The angle of each canthus is measured. I use a protractor to do this. Its accuracy depends on whether both people in the photos you're comparing are looking ahead normally. If one is wide-eyed and the other is obviously squinting, it is of little value.
Q. The "preliminary" analyses in your articles seem pretty flimsy to me. How can you say for sure that a photo is or is not someone after such a quick glance?
A. I can't and I don't. The preliminary analysis is more than enough to rule out matches, and even if I think a face is not a match based on sight alone, I still measure the faces to ensure accuracy. All of the photos mentioned in my articles about face biometrics have been through the preliminary analysis and were also measured down to 1/24 of an inch.
Q. Can you explain why you couldn't compare the faces in my painting and photograph?
A. I did compare them and they are not the same person. In fact, they do not resemble each other at all. It is a waste of your money and my time to measure faces that are obviously not the same. I can tell you why they are different from a glance. The ears don't match, the nose in the painting is longer, the eyes are farther apart in the photo, the forehead is substantially lower in the painting, the hairlines are different, the lips are different, and I could go on and on and on. I can't find anything that is the same in the two faces.
Q. Do you need to be a physical anthropologist or a forensic anthropologist to be a facial features analyst?
A. No. Physical anthropology (bioanthropology) is about human biology and how humans came into being, how we adapted and changed over time, particularly in relation to our non-human relatives, the primates. Forensic anthropologists, on the other hand, deal mainly with bones, usually of people long dead and already turned to bone, and they are known most popularly these days for their facial reconstruction skills. It would be rare to find a physical anthropologist or forensic anthropologist who would be educated sufficiently in face biometrics to authenticate identities of people in photographs. It is a specialty unto itself and is pretty far out of their areas of expertise.
Q. Can you analyze facial features if a person is facing straight ahead in one photo and the other person is in three-quarter view?
A. Usually, yes; sometimes, no. It depends on the individual photos and how much is visible in each to make a comparison. If both photos are very clear with a lot of detail, then the chance of successfully comparing the two different views is much better, especially with regard to the vertical line-up of the features.
Q. In your book you mention the angle of the head when making vertical measurements. How can you tell for sure if a head is tilting up or down? Can a photo editor be used to straighten such an alignment?
A. It is actually fairly simple to determine these angles. When a head is tilting down, more of the top of the head above the trichion (hairline) is exposed, and less of the gnathion (the lowest part of the chin or bottom chin edge) is visible. When a head is tilted upwards, less of the top of head is visible above the trichion, and the underside of the nose (generally at least a part of the columnella) is exposed. This type of angle cannot and should not be corrected as it would likely create inaccuracies in the measurements.
Q. I have a copy of your 36-page 1992 booklet, Anthropometry and the Human Face in Photographs. It doesn't mention the latest ways to analyze faces. What do you suggest?
A. Thanks for waiting. The greatly expanded 192-page version is now available. Its new title is Face To Face: Analysis And Comparison Of Facial Features To Authenticate Identities of People In Photographs.
Q. You seem very negative towards historians who authenticate photos. I would think they are the best sources for accurate authentication.
A. Historians are often very good at authenticating photos. But I don't authenticate photos. I am an anthropometrist and I authenticate identities of people in photos, and I use my knowledge of cranial and facial and ear anatomy to do that. I don't rely on provenance at all. The fact that a photo was found with other photos of the same person or relatives or associates of that same person, or that the photo was taken in a particular studio is irrelevant to what I do. I let the person's features tell who they are. Most historians don't do that very well, if at all.
Q. When you measure the distance between facial features on each face, how do you then compare those measurements? It would seem they would always be different because the sizes of photos are different.
A. You are absolutely correct. You compare the percentages of the face that compose the sizes of the features in proportion to the rest of the face. For example, if the height of the chin in one photo represents 19% of the total length/height of the face, then you would want to find that same percentage in the chin of the face to which you are comparing it for that feature to be a match.
Q. I have an old photo of my great-great-uncle who is Norwegian. The photo is kind of small, but you can see his hooded eyelids. I don't think we have any Asian ancestors. What else could account for this?
A. He may just have heavily hooded eyes and not the true epicanthic folds found in the eyes of some indigenous Asian peoples. If that's not the case, perhaps he inherited the trait from a much earlier Saami (Sámi) ancestor.
Q. How many vertical bone measurement ratios do you use to determine that any two faces are not a match?
A. Technically, only one is necessary to create reasonable doubt. But, it depends on how far off that measurement ratio is from a match. To be on the safe side, I always do all the vertical measurements before reporting my findings, and I check my measurements and my ratios three times, because accuracy is so critical to making a finding.
Q. How important is a photo's provenance when identifying people?
A. As an anthropometrist specializing in facial features, it's not important to me at all. If a face is an exact match for the person it's purported to be, then the provenance merely matches that person. Where and when a photo is taken is of interest in dating the photo, but it won't make a person be who you think it is or want it to be unless the face is analyzed and compared with a known photo of that person and is found to be an exact match.
Q. Why don't you compare photos of children with photos of their adult selves? Can't age progression software make this possible?
A. I don't do age progression. It isn't my specialty. It requires a lot of research into family photos to help determine how a person might age based on how their ancestors did. Even then, the result is just a guideline. In that way, it is a great tool for assisting law enforcement in finding missing children who have grown up, but it would not be adequate for making an exact match of a child's and adult's photo.
Q. I'm trying to learn how you do what you do so that I can identify some of the hundreds of family photos I have. One of the problems I keep coming up against is being able to make a good comparison of the lips. They all seem to look so similar.
A. That's probably because they are part of the genetic resemblance in your family photos. I've identified a lot of people in family albums, and it is very challenging to do so. All I can say is that as you keep studying your photos, you will eventually be better at catching the very subtle differences that distinguish one family member's lips and their other features from another family member's similar features.
Q. You mention in an article that the alignment of facial features with the ears might not be useful if a photo is taken at an angle. Why?
A. If the head is tilted upward or downward in a photograph, the ears will not align the same way with the features on a face that is looking straight into the camera.
Q. I have several pictures of my great-grandmother, and her weight varies throughout her life. Are there any tricks to analyzing these photos to be sure they are all of her?
A. In most cases, weight gain/loss doesn't make much difference in analyzing features, especially the ears, and also in a vertical comparison of the facial features. The overall shape of the head and face may change slightly with weight gain, particularly the zygomatic span (cheek to cheek), but the ear placement and configuration, the eye size and shape, pupillary and other ocular distances, and possibly the nose and lips, will remain the same, as will the alignment of all the other features that rely on underlying bone structure.
Q. I have two Daguerreotypes that I was told are of my great-great-grandfather who was a slave in Mississippi. But he is very dark in the Dags and our family history says he was half white. In another photo of him, a CDV taken in 1887 when he is much older, he is much lighter, even though he looks like the same person. Could those old Dags be of someone else, a brother perhaps? Any thoughts on this?
A. It is certainly possible that the Daguerreotypes are not of your great-great-grandfather. But, Dags had very long exposure times, and as a result, the final image was very high contrast – what was dark was darker and what was light was lighter. The later methods of photography didn't require that long exposure time, so the contrast was not as high and provided a more natural appearance. So they could all be photos of your great-great-grandfather, differences in skin tones courtesy of the photographic processes.
Q. Do you use any kind of facial reconstruction software that can help identify a face that was severely damaged in an accident?
A. Sorry, that's not my specialty at all, so I don't know what kind of software might be available for it.