J. Cosmet. Sci., 60, 111–123 (March/April 2009) 111 The shine problem in hair: Review of imaging methods and measures for luster P. D. KAPLAN, K. PARK, J. QI, and K. YANG, TRI/Princeton Synopsis There is a need for both a better understanding of the technical drivers of shiny appearance in hair and for standard methods and measures of shine. To this end, we develop standard treatment methods for changing hair shine and examine a number of image-based measurements of luster. Using psychophysical techniques to get a perceptual reference, we fi nd that available technical measures are diffi cult to use when trying to quantify the small changes in shine associated with treatment. INTRODUCTION Despite its intuitive simplicity (1,2), and years of research (2,3), the technical measure- ment of shine remains problematic (4–6). In addition to the technical challenges presented by complex illumination, heterogeneous materials, shape, and roughness, the very nature of the “shine” question can be diffi cult to describe even when the question is asked within a single restricted application. One way to illustrate the complexity of the question is to see how the question itself changes radically with context. For example, the computer graphics professional needs an answer which will enable the production of a realistic image (2), while the vision neuroscientist hunts for biologically realistic computations, essen- tially image analysis algorithms, that could possibly be similar to those taking place in the mind of an observer the coating chemist requires answers which will help in the engineer- ing of a surface that will produce, in the observer, a favorable qualitative judgment. While the shine question for each of these practitioners is related, a successful measurement for the computer graphics community may not be at all useful for the cosmetic scientist. As cosmetic scientists, we are concerned with the often subtle, optical changes produced by thin, nearly transparent layers of particles and fi lms. In this work we restrict our inter- est to the shine of straight hair. Two recent advances motivate us to undertake this work. The fi rst is an improvement in the form of a commercially available, special-purpose polarization-imaging system that should enable computation of shine without requiring much signal processing (7). Second, there is an opportunity to extend recent progress by neuroscientists on the nature of shine to hair as a specifi c substrate. We fi nd both advances to be intriguing and to advance the state of the art, but we also fi nd through the experimental results presented here, that the goal of capturing essence of shine in hair using instrumental techniques remains elusive.

JOURNAL OF COSMETIC SCIENCE 112 Polarized imaging provides a straightforward route to separate specular refl ections from diffuse refl ection because the diffusely refl ected photons are, at least in the limit of mul- tiple scattering, unpolarized. As a result, a pair of images, one formed by photons with polarization parallel to the illumination I|| and one perpendicular I⊥ can be used to sepa- rate the specular, shine image Is = I|| − I⊥ and the diffuse image Id = I⊥. The formulas published in the literature can then be computed without detailed computations on the angle-dependent refl ection curve. We have developed both positive and negative product controls using this imaging system and present an analysis of the sensitivity of the tech- nique for measuring distinctions between hair following different treatments. The new insight from neuroscience uses a simply-computed quantity, the skew of an im- age intensity histogram, as a technical surrogate for shine. Beyond computational sim- plicity, this approach reports to be somewhat color independent, a property that could be particularly useful, given the strong lightness dependence of the currently published technical measures of shine (8) Skew, mathematically the third moment of a distribution, is a measure of asymmetry about the mean. The underlying concept is that one might expect image histograms to be more or less normally distributed in the absence of shine but the pixels in which shine is visible should be far brighter than the rest. Rather than fi nding peaks and analyzing the intensity, or the counting shiny, skew incorporates both of intensity and extent in a rationally weighted way that has the virtue of being a neuraly plausible computation for the visual system. Brighter pixels count more, and every pixel brighter than the mean that is not balanced by a similar pixel below the mean contributes to the skew. We examine the dependence of skew and other measures such as the Reich- Robbins formula (4) and the TRI formula (9) on the lightness (L* in L*a*b* space) of hair and fi nd this measure to be intriguing. One key test of any technical measurement is its ability to predict perceptual judgments. We subject all of these measurements to this test by creating a series of hair damage standards and checking by the perceived differences in shine and measuring technical differences. Our fi nding is that the perceptual differences are clear and consistent, but that none of the technical measures closely follow the pattern of perception for small changes in shine. MATERIALS, METHODS AND MEASURES HAIR All hair used in this work is blended, straight European hair. A list of colors and their measures are found in Table I. Prior to treatment or damage, hair was pre-treated with a cleansing shampoo. CARE METHODS For a simple control, a low dose of phenyltrimethicone (0.0125g/g hair) was applied to the hair tress, spread uniformly down the tress by gentle kneading and air dried. For the conditioner and 2-in-1 formulations, the hair, was pre-wet, treated with 0.15g product/g hair, massaged for 30 seconds and rinsed for 30 seconds. Water supplied at 37°C, 1.6 gallons/minute.