Seeing Is Not Believing: Facial Distortion in Smartphone Photography

Smartphones are ubiquitous, and their ever-improving cameras make photography more accessible to the public and to surgeons.1 In 2014, more than 93 billion selfies were taken per day, and the social media photography platform Instagram reached 1 billion users in 2018.2 This increased exposure has influenced how users assess their own appearance and has increased demand for aesthetic procedures.3 In addition, the convenience of smartphone photography has changed practice for surgeons, who use smartphones for office photography and social media.1,4 Unlike traditional portraiture, smartphone photographs are taken with the subject very close to the camera. This necessitates wide-angle lenses with short focal lengths, which allow better capture at close distances. The optics of wide-angle lenses produce relative magnification of the central portion of an image, at its extreme creating a fish-eye effect.5 It is important to recognize the degree of distortion present in smartphone photographs, as this could impact a patient’s self-image and how surgeons document preoperative and postoperative results. To systematically study facial distortions in smartphone photographs, we compared the front and rear cameras of an iPhone 8 (iOS 11.0; Apple, Inc., Cupertino, Calif.) and a Galaxy S7 (Marshmallow 6.0.1; Samsung, Seoul, Republic of Korea) to a professional digital single-lens reflex camera [Nikon D800 (Nikon Corp., Tokyo, Japan) with a 50-mm fixed lens]. We photographed a live model and a geometric grid (56 × 72-cm, 4-cm2 boxes) in a single studio session, as directed by a professor of photography. [See Figure, Supplemental Digital Content 1, which shows an overview of photographic imaging setup. A boom system was used to position the cameras for the duration of imaging. Three photographs were taken at three positions relative to a horizontal line from the profile of the model’s nose/center of the grid: level (0 degrees from horizontal), above (30 degrees above horizontal), and below (30 degrees below horizontal). At each position, the camera was placed 75 cm from the model’s face to reflect average arm length, https://links.lww.com/PRS/E622.] Using digital single-lens reflex–captured images as our standard, we evaluated distortion with cephalometric ratios or, for the geometric grid, height and width ratios. We expected a fish-eye effect for the smartphones, given their shorter focal lengths, but photographs of the grid did not differ from digital single-lens reflex images (p > 0.05). [See Table, Supplemental Digital Content 2, which shows the comparison of vertical and horizontal peripheral to central measure ratios obtained from images of geometric grids for smartphones compared to digital single-lens reflex cameras. Difference between ratios demonstrated as percentage increased or decreased ratio at stated photograph angle (p < 0.05) as determined by two-tailed t tests, https://links.lww.com/PRS/E623.] Interestingly, faces had inconsistent distortions, where sometimes the central face was compressed and sometimes it was magnified (Fig. 1). Differences in distortions within the same camera indicate that distortion was predominantly a result of image correction software. There are relatively simple algorithms for correction of lens distortion, and this likely explains the concordance between cameras when photographing the grid, but these algorithms often produce noticeable artifacts in the periphery of photographs (e.g., curved edges on buildings).5 Now, more sophisticated processing, based on artificial intelligence and facial recognition, attempts to automatically “correct” facial distortions while preserving the background.5Fig. 1.: Visual side-by-side comparison of unedited 0-degree photographs from iPhone rear camera and Samsung rear camera. Cephalometric ratios obtained for smartphones compared to digital single-lens reflex cameras for nasal height/facial height; nasal width/facial width; and interpupillary distance (IPD)/facial width. Difference between ratios demonstrated as percentage increased or decreased ratio at stated photograph angle (*p < 0.05). NS, not significant.Our results indicate that corrective algorithms in the smartphones tested were not robust enough to account for the photograph angle, or even differences between front and rear cameras. Troublingly, image interpretation and correction are moving targets, as hardware and software continually change with upgrades for both native camera applications on smartphones and third-party applications. Smartphone photography’s unpredictable distortions can change patients’ self-perceptions and introduce inconsistency when used for documentation. Our study highlights key concerns with smartphone photography for medical documentation, especially for facial plastic surgery. Based on our findings, we suggest that plastic surgeons should be equipped with a standardized digital single-lens reflex photography setup. This would ensure accuracy and consistency in medical photography for patient documentation, social media, quality improvement, and research purposes alike. PHOTOGRAPHIC CONSENT The subject provided written consent for the use of her images. ACKNOWLEDGMENT The authors would like to thank Adrian Fish, M.F.A., for help with this project. DISCLOSURE The authors have no financial interest to declare in relation to the content of this article.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.