Animated emoji in augmented reality (AR) enables users to create a humanoid version of themselves that mimics their facial expressions dynamically. In this study, we aim to explore how people perceive facial skin color in digital portrait in comparison with humanoid emoji in AR. We tried to identify the skin color representative regions and to estimate the color difference between the two contexts. We conducted a user study comprised of three tasks with 20 graduate students majoring in design and employed 24 portrait images in four skin tone categories. Through the user study, we first figured out that forehead and cheek regions, and particularly the linking band between eye and lip, were often considered as the representative region of facial skin color. Second, we observed skin colors become lighter in general, except dark tone. Furthermore, concerning the vidvidness, all four skin tone types became paler in humanoid emoji. Diverse ethnicities and contexts are expected in the future to provide a more robust and reliable analysis of the perception of skin color.
Regular observation and recording of the changes in body appearance are essential for the process of the treatment of plastic surgery and dermatology, especially aesthetic surgery. Usually, physicians treat patients with medical interviews, pictures of the patient's faces before and after their treatment, anatomical data that including size, location, and color of the affected skin. However, it is difficult to capture the affected area under the same conditions every time because the captured range varies depending on the imaging angle and distance. There is a need to record three-dimensional shape of face parts such as cheek, nose, eye, and chin. Therefore, in this study, the face shape and the skin color were measured using the infrared depth camera and the RGB camera built in the smartphone three-dimensionally. We measured before and after modulating the shape and color of the face, and then, the change in volume and the change in skin pigment of skin color was calculated and visualized. This method makes it possible to analyze the skin shape and color independently of the viewing angle and the illumination direction. In this study, the depth sensor built in the smartphone showed the potential to monitor changes in facial shape and skin color. In the future, it is expected to contribute to the development of telemedicine, in which the patient measures their face at home and gets medical treatment consultation remotely.
This research examined the performance of skin coloredpatches for accurately estimating human skin color. More than 300 facial images of Korean females were taken with a digital singlelens reflex camera (Canon 550D) while each was holding the X-Rite Digital ColorChecker® semi-gloss target. The color checker consisted of 140 color patches, including the 14 skin-colored ones. As the ground truth, the CIE 1976 L*a* b* values of seven spots in each face were measured with a spectrophotometer. For an examination, three sets of calibration targets were compared, and each set consisted of the whole 140 patches, 24 standard color patches and 14 skin-colored patches. Consequently, three sets of estimated skin colors were obtained, and the errors from the ground truth were calculated through the square root of the sum of squared differences (ΔE). The results show that the error of color correction using the 14 skin-colored patches was significantly smaller (average ΔE = 8.58, SD = 3.89) than errors of correction using the other two sets of color patches. The study provides evidence that the skin-colored patches support more accurate estimations of skin colors. It is expected that the skin-colored patches will perform as a new standard calibration target for skin-related image calibration.
This research suggests a color constancy algorithm using the human sclera and pupil for estimating skin color. The human sclera is approximately white, reflecting the hue characteristics of the illuminant. On the contrary, the pupil is shown as approximately black, independently of the surroundings. Consequently, the sclera and pupil account for the color characteristics of a facial image. Based on this assumption, a new color constancy algorithm was developed, and we examined the performance by comparing the calibrated colors with actual skin colors measured with a spectrophotometer. As the dataset, we collected facial images as well as the CIEL*a*b* values of 348 Korean females. As a result, the error of the proposed algorithm was significantly smaller than the state-of-art skin estimation algorithms. The algorithm developed in this study provides evidence that the human sclera and pupil successfully serve as the calibration targets to estimate the color of human skin.
The determination of local components in human skin from in vivo spectral reflectance measurements is crucial for medical applications, especially for aiding the diagnostic of skin diseases. Hyperspectral imaging is a convenient technique since one spectrum is acquired in each pixel of the image, and by inverting a light scattering model, we can retrieve the concentrations of skin components in each pixel. The good performance of the method presented in this article comes from both the imaging system and the model. The hyperspectral camera that we conceived uses polarizing filters in order to remove gloss effects generated by the stratum corneum; it provides a high-resolution image (1120 × 900 pixels), with a thin spectral sampling of 10 nm over the visible spectrum. The acquisition time of 2 seconds is short enough to prevent movement effects of the imaged area, which is usually the main issue in hyperspectral imaging. The model relies on a two-layer model for the skin, and the Kubelka–Munk theory with Saunderson correction for the light reflection. An optimization method enables computing, in less than one hour, several skin parameters in each of the million of pixels. These parameters (blood, melanin and bilirubin volume fractions, oxygen saturation…) are then displayed under the form of density images. Different skin structures, such as veins, blood capillaries, hematoma or pigmented spots, can be highlighted. The deviation between the measured spectrum and the one computed from the fitted parameters is evaluated in each pixel. © 2016 Society for Imaging Science and Technology.