JOURNAL OF COSMETIC SCIENCE 320 rested on a hard stand at a certain angle. A white light source (14 W) was placed at a distance of 24 inches from the mannequin’s head. The digital camera was positioned at a distance of 12 inches, providing an angle of approximately 30° relative to the central axis of the mannequin’s head. The digital camera was mounted on a stationary mount and connected to a computer. A color code chart was placed at the base of the manne- quin in order to calculate the white balance and provide a scale. A Cannon EOS 20D digital camera with a resolution of 8.2 MP, and equipped with an EFS 17-55 mm (f/3.5-7.1) lens, was employed as the image collection device for all studies presented in this article. Digital images were captured as raw images fi les (CR2) remotely by a computer. The white balance was adjusted to the corresponding lamp color temperature, 3800°K (Cam- era Raw, Photoshop 10 – Adobe Systems Inc). Image analysis was carried out using Im- ageJ version 1.42q software (NIH), which enabled us to obtain light intensity (luminance) distributions along the lower lip of the mannequin. A 950 × 550 pixel image of the lip area of the mannequin was cropped from the original image the shine band of the lower lip was further cropped from this image and analyzed for luster. The image was rotated 90° to the right, and converted to 8-bit, and a profi le was generated measuring the light intensity across the distribution gradient. The analysis was applied to all images of all treatments. Luster calculations. The occurrence of refl ection is the result of the interaction of light with a substrate, based on its material properties. The interaction of light with objects also creates scattering, refraction, diffraction, interference, and adsorption. The most important part of luster comes from the specular and diffuse refl ection. Luster measurements of hair have been the subject of research for the past 35 years, and are unlike the luster measurements of lips, which have been mostly qualitative (4). A goniophotometer was employed by Stamm and coworkers in 1977 (5) to record the light distribution curves necessary to calculate the luster parameters of hair. Fur- ther work with goniophotometry was completed by Reich and Robbins (6), who were able to show a correlation between this quantitative technique and qualitative con- sumer studies. McMullen and Jachowicz (7) employed image analysis to calculate the luster parameters of hair by utilizing high-resolution digital photography in con- junction with the data analysis procedures set forth by the goniophotometric tech- nique. The equations used to calculate luster were adopted from Stamm et al. (5) and Reich and Robbins (6). Figure 1. Diagram showing the positioning of the mannequin in relation to the light source and the digital camera.

EVALUATION OF LIPSTICK LUSTER 321 Stamm et al. (5) proposed a luster parameter based on the specular and diffuse refl ectance area under the curve: - = Stamm S D L S (1) Reich and Robbins (6) offered a slightly different relationship to defi ne luster: R S L D W1/2 -R = ´ (2) where S represents the area underneath the specular curve, D represents the area under- neath the diffuse curve, W1/2 represents the width of the specular peak at half of its max- imum intensity, and L represents luster or shine. A schematic describing the specular and diffuse refl ectance is presented in Figure 2. Statistical analysis. Normality of the data was checked using the Wilk-Shapiro test, and equality of variance was checked using Bartlet’s test. When normality failed, a Kruskall- Wallis non-parametric evaluation was performed along with a Dunnett’s test to compare multiple treatments to a control. When equality of variance failed, a Dunn’s test was performed to compare multiple treatments to a control. Statistical analysis was performed using Sigma Plot 11 software (Systat Software, Inc., San Jose, CA). RESULTS Figure 3 provides digital photographs of treated lips with the formulations tested, the cropped photographs with the shine bands, and the converted 8-bit gray scale images. It is very noticeable from the images that differences among treatments are more noticeable in the cropped image than in the full-lip images however, gray scale images accentuate the differences the most. Based on visual observations, it can be seen that all treatments had more shine than the control, with treatment D (5% polybutene) being the least dif- ferent and treatments C (7% VP/hexadecene copolymer and 4% VP/eicosene copolymer), Figure 2. Schematic of the calculation of the luster parameter, which shows the difference between specular and diffuse refl ectance.