Practical Implications For Filmmakers, Photographers, Stereographers & Camera Designers
Despite countless publications and experiments involving body image, BID in photography had not been systematically investigated until these experiments. And until recently, it had not been recognised as a significant effect by any other resource or authority on professional imaging. However, the gender bias found experimentally has been reproduced for the first time (linked image above) by the distinguished cinematographer and renowned authority on digital imaging, Geoff Boyle NSC FBKS. He used professional previsualisation CGI software to show that when the image size is kept constant and the virtual camera to subject distance is increased, even the finest lenses will change the apparent size and shape of virtual humans.*
*This author estimates that the female figure in the CGI animation has a virtual Body Mass Index below 18 BMI. If so she would be a perfect substitute for many leading actresses and top models, (who are often below the healthy 18-21 BMI range) and have deliberately reduced their body fat levels to improve their photographic image. However if they are interpreted as a virtual example of a normal female, the CGI animation will be misleading. The negative effects of focal length changes on attractiveness and apparent body weight are likely to be much more apparent for females if the subject has a slightly higher BMI in reality (as seen in the homepage photographs). This may encourage dangerously low BMI women to lose weight to appear more attractive in photographs.
This Body Image Distortion research is the only guide photographers can use to quantify the real-world effects of focal length on body image and gender. The graph below reveals the dramatic fattening effects of telephoto lenses on apparent female body weight. Males under the same lighting and with the same camera to subjects distances had no fattening, or increase in perceived weight.
The graph shows rising weigh estimations in kilograms as the camera to subject distance increases. All of the female weight estimations are seen as dramatically overweight (above the median line). The males however are never fattened under the same conditions.
Photographers usually discover BID in photography without any theoretical or practical assistance (unless they were guided by a exceptionally experienced hand). However, this Body Image Distortion in Photography research should provide overwhelming evidence to even the most sceptical of photographers: 2D photography is predictably and significantly distorting when photographing all people, but is innately fattening only to females. The graph above is from an experiment using 6 male and 6 female models. The groups included underweight and overweight participants, but the group-average Body Mass Index was the same as the general population.
Only orthostereoscopic images presented life-size can convey accurate size, shape and distance information. This is why even the most experienced photographers can be surprised at how capturing 3D reality in 2D can fail to reproduce what they saw with their direct vision. They may then look for the cause in the wrong places, such as “wrong” lenses, sensors, actors, poor lighting, settings, post-production issues etc. Lack of knowledge of how significant BID and other variables can interact will inevitably lead them into errors and avoidable re-shoots.
Experienced photographers however will naturally develop analytical vision when assessing
real targets and their subsequent 2D rendering. Experience allows them to predict how the true colours, size, shape, height, weight and attractiveness of the real scene might be reproduced, even though it could be camouflaged or illuminated in ways not conducive easy visual assessments. BID in this context is just one of many more widely understood features of real world photography that reproduce differently to reality. As such, BID is one of the many variables photographers have to control for to have mastery of their professional practice.
Body Image Distortion is the predictable and inevitable consequence of choices made during the image-making process.
A practical application of these experiments is to introduce inexperienced image makers and media professionals to a set of variables they are probably completely unaware of. In these sections advice is given for photographers & stereographers, producers, set designers, hair & makeup and on-screen talent that is unavailable elsewhere. For highly experienced photographers these experiments should confirm some of what they have known for decades. Most importantly, the effects reported here are highly significant both experimentally and in professional practice. For instance, many media professionals are aware that models and on-screen talent can be fattened by the imaging process. So they cast for slimness, while photographers and stylists use many other interventions such as hair, makeup, lighting, clothes and posture to compensate for these variables. But without a theoretical understanding of the variables, mistakes are not only possible but in some instances may be the rule rather than the exception.
For example, most professionals consulted during these studies were very surprised to discover that slightly wide-angle lenses are both slimming and have no detectable distortions in side-by-side comparisons. Even the “balloon face” distortions created by extreme wide angle lenses were not interpreted as fattening in any of these experiments. The reason is that human vision is curvilinear by nature so can interpret near fish-eye photography. But most importantly, the reduction in neck width seen in wide angle portraiture is always perceived as slimming. Further, any increase in the apparent size of eyes relative to the rest of the face is also powerfully suggestive of youth and attractiveness. So if people are looking slightly larger, older or fatter on the screen than in reality (or is required for the shot), selecting a wider angle lens can be beneficial in correcting these effects.
The 3D experiment summaries are below. But the take-home message for stereographers is that it confirms many of their core beliefs: 3D images can be much more naturalistic and life-like, with far more perceptually accurate and useful information than can be conveyed by 2D. Stereographers have often noticed that 3D is much sharper too. This is now attributed to a perceptual process called binocular summation, which is how the brain allows stereo pairs to convey 1.4 times more resolution than can be seen in either channel viewed in 2D. Changing Inter-Ocular Distances will change apparent size, with the most accurate size estimations conveyed by converged 60-65mm IOD images viewed life-size. Smaller IODs give larger size estimations, larger IODs give smaller size estimations (leading to obvious dwarfism even without side-by-side scale comparators).
But stereographers and filmmakers are still completely unaware of the subliminal effect of 3D on mood and emotional responses. Research by Bernhard Goodwin and Hannah Freuh of the Department of Communication Studies and Media Research (IfKW) at LMU Munich has shown that 3D has remarkable behavioural effects. In a study looking at aggressive and negative emotional responses to horror movies, they discovered an opposite effect to what they were expecting. As predicted, 3D horror movies were significantly more immersive and involving than otherwise identical 2D presentations. But the post-hoc analysis of attitudes to the film were remarkable: 3D audiences were significantly more satisfied and calmed by the presentation mode than 2D-only viewers. The positive effects of 3D on the audience are likely to suppress post-presentation adverse emotional reactions, so 3D was less likely to have harmful emotional effects in the general population. When asked if the effect was big enough to make it almost unethical to show a 2D version of a horror movie made in 3D, the author said: “I think the unethical claim is too strong. But I would say the industry should be aware of it.”
Why is 3D still so misunderstood? Stereography is almost as old a photography itself. But binocular vision is a complex perceptual process that has only recently been open to detailed neurophysical study. One newly discovered aspect of human vision is that it actually relies on four separate visual channels to send optic flow, spatial, cognitive and colour information to the stereo cortex of the brain. Yet when all four channels are working perfectly, we do not “see” directly through any of them! This counter-intuitive finding is attributed to a number of neural processes that fuse the left and right image channels to form a cyclopean view. When we fuse two optical channels into a single cyclopean viewpoint, the resulting percept is fundamentally different to a 2D photo-mechanical image. It is measurably superior to monocular viewing, though its many similarities to 2D imaging has allowed us to believe it is perceptually almost identical. The research reported here is just one of many rarely reported experiments that may explain why 3D imaging is so often misunderstood even by vision scientists, professional stereographers and image makers using state-of-the-art equipment.
Camera designers never consider a simple fact: Compressing the Z-axis of infinitely deep 3D space into an infinitely small 2D plane can have adverse perceptual consequences. In the past, photographers could argue about whether the camera really did add 10 pounds to apparent bodyweight or if it was just a myth believed only by overweight actors. Now they cannot. We know the fattening effect is real and can be larger than 10% of body weight. Worst still, we also know that the loss of Z-axis depth produces gender-biased fattening, with women being very adversely affected by the flattening effects of 2D imaging. Yet 2D is the default method of viewing photography for almost every audience and in almost every media outlet. Even If portraits and close-ups in movies were taken by a 3D camera, only a tiny percentage of the audience will ever see them in native 3D and be able to see a more natural reproduction of body image.
For the future of camera design, it is important to remember the pursuit of more resolution, perfect colour and tonal reproduction is not enough. The human visual system uses cyclopean perception to see colour, tonal range, shape and motion with two eyes that have far more information gathering capacity than any 2D camera. Naturalness in photography is not defined by sensors, lenses or software but by how much useful information can be gathered and how naturally it can be reproduced. After almost 200 years of research and development, is seems that the most natural photographic process is still beyond our grasp. So even the best photography always looks like a mediated reproduction of reality, rather than a truly life-like image. But for how long will that be acceptable, or allowed to be true?
The Human Visual System is Foveal,* Transsaccadic* and Cyclopean* These are core perceptual processes discovered by 20th century vision scientists that have never been applied in commercial image making. Stereoscopic 3D can be highly compatible with these processes, but has often been dismissed as a gimmick by media professionals, or as inferior to 2D as an artistic mode of expression. Yet in the long history of research into stereoscopic vision, it is a remarkable fact that the superiority of 3D compared to 2D is overwhelming. The Goodwin & Freuh research (above) supported by Arri on emotional and behavioural responses to 3D movies has also shown that it too can be far superior to 2D, with a number of very surprising findings. Behavioural studies like this concur with the history of psychophysical experiments and the detail analysis of the BID research. Though some results still have the power to surprise, such as the finding of almost no correlation between eye-strain and IOD (reported in bold below).
Binocular vision is measurably superior to monocular viewing: 3D offers a 40% increase in resolution and a visible improvement in signal to noise ratio while viewing any type of film or video image. 3D also improves size, shape and distance perception, field of view, motion perception, colour discrimination, luminance increment detection, contrast sensitivity, vernier acuity, visual search, letter identification and almost every other measurable feature of the human visual system. These factors are in addition to our ability to extract useful information from each of five separate ocular vergences (delivered through the four separate perceptual channels). The stereo cortex then delivers an image to the brain that has a vastly more detail and information value than is possible from a single eye, or from a 2D lens image of the same magnification. Stereoscopic hyperacuity and cyclopean vision are the keys to unlocking why 3D movies have yet reach to reach their theoretical potential. Hopefully the research detailed on this website (stereoscopic summary below) will lead to improvements in camera design and photography that are currently near the maximum possible deliverable by single lens 2D imaging. When reviewing the many positive findings from the history of vision research, it is becoming harder to understand why 2D is still considered to be “better” than 3D.
When comparing monoscopic 2D and orthostereoscopic 3D images, the geometrically accurate stereoscopic image is always perceived as slimmer than otherwise identical monocular stimuli. The 3D images of people correlated to their actual Body Mass Index, whereas the 2D images always correlated to a heavier body mass. This strongly suggests that the 2D condition is the distorted condition.
The slimming effect of orthostereoscopic imaging was observed in photographs of people, non-human objects and computer generated images.
The slimming effect of binocular disparity follows a dose-response trend: as larger stereoscopic disparities (camera separatations) were projected, progressively slimmer object sizes were observed.
All participants were tested for stereoscopic vision using the TNO and Frisby tests. 8% of subjects had zero stereo-acuity scores on the TNO (red -green anaglyph) test, a clinical indication of poor to zero stereo vision. Yet they saw full depth images and had data indistinguishable from people with the highest scores. This suggests anaglyph 3D is extremely poor for 3D reproduction, as the same subjects viewing the natural depth Frisby Test could achieve scores near to or above average.
Zero eye-strain is a feature of convergence stereography. Despite using stereo bases that were much wider than in modern commercial stereoscopic practice (30mm-240mm of IPD), there were no reports of eyestrain at the 30, 60, 63 and 120mm conditions. Only at wider separations were there occasional reports of discomfort, but these were always associated with those people who had lower than average Frisby test scores. The conclusion was that eye-strain is not an issue with orthostereoscopic imagery when it is reproduced life size and with all natural vergences accurately reproduced on the screen.
* For more information on Cyclopan Vision, Foveal Imaging or Transsaccadic Perception and pdfs of the research, please contact email@example.com