Study on Skin Wrinkles and Pigmentation Analysis Methods Using Multispectral Images

Abstract

The attention to skin care has been steadily maintained in recent decades. Lots of men started to care for their skin and LED masks became popular for therapy at home. After getting a treatment in a clinic or using products such as LED masks, people simply observe their appearance to confirm the improvement of skin condition. However, it is hard to notice the changes in appearance because skin treatments take a long time. In the hospital, specialized devices are used to measure skin conditions for diagnosis. Most are expensive and large for home uses. As a first step to develop a cost-effective, portable, and multifunctional skin diagnostic device, we studied a basic principle of measurement methods using visible and near-infrared images. Especially, we focused on the wrinkle, elasticity, melanin, and erythema. As time passes, skin becomes less elastic and facial wrinkles are developed. Wrinkles are generally classified as the wrinkle index by the appearance. Most people can easily understand the overall degree of wrinkles because the wrinkle index is a simple and intuitive method. However, the diagnosis may be affected by the doctor’s subjective experience and the result can be different even with the same skin condition. The other method is an image reconstruction. Although it provides rich information, the surface of the skin is replicated and the replica is reconstructed to images with a heavy computational load. Therefore, it needs a simple and objective method for non-expert people to understand the overall wrinkle condition intuitively. Similar to the skin wrinkles, fabric wrinkles have been studied to be classified into grades by calculating the entropy of the Fourier transformed image. The brightness is suddenly changed near the wrinkle lines and it makes high frequency components. As the degree of wrinkles increases, high frequency components increases and it leads to an increase in the randomness of frequency components. This randomness is quantified by calculating entropy. We applied this entropy method to skin images to evaluate skin wrinkles. Eight sets of skin images are used and each set consists of six images with the wrinkle index from grade 0 to 5. Entropy values are divided by the ‘grade 0’ value and a tendency that how much the entropy is changed with increasing the wrinkle index is analyzed. The results showed that the normalized entropy increases as the wrinkle index increases, especially for lower index. Skin wrinkles can be classified into shallow and deep wrinkles according to their cause and formed skin layer. Reduced skin elasticity causes wrinkles in the dermis layer, where near-infrared light can penetrate into. In this study, the relation between skin elasticity and the entropy of near-infrared skin images has been analyzed. From 12 Korean adults, cheeks and forehead were measured using a clinical device ‘Cutometer’ and skin images (470 nm, 870 nm, and broadband light) were acquired. In the results, the measured data of the 870 nm image is close to the fitted model and shows the tendency that entropy increases as skin elasticity decreases. Furthermore, the results are classified by sex, age, and measurement area and it shows the same tendency with the previous studies. For melanin and erythema, ‘Diffey algorithm’ was applied to skin images to see the possibility that the imaging method could replace the expensive commercial device. This method was verified by the phantom test and simulated images. After verification, a total 16 measurement data was acquired in the hospital with a commercial device ‘Mexameter’ and imaging system. As the measured value of Mexameter increases, the index calculated from images increases. The correlation coefficients of melanin (0.725) and erythema (0.810) are not large enough to use the imaging system instead of the commercial device. In addition, melanin distribution was acquired by using wavelength-dependent light properties. The distribution of areas can be affected by both the density and depth of melanin. In this study, skin wrinkles are evaluated by applying the entropy methods, which have been studied in fabrics. The relation between skin elasticity and the entropy of skin images were analyzed by applying this method to visible and near-infrared images. To see the possibility of replacing the clinical device to the imaging method, the melanin and erythema indices were analyzed. This study showed the possibility of developing home-use devices to measure wrinkle, elasticity, and pigmentation in the future.