New spectral data for skin colours

Two experiments were conducted for collecting skin database at the Universities of Liverpool and Leeds (UK), and Zhejiang University (China). Overall, 235 subjects from 4 different skin groups (Caucasians, Chinese, South Asian and Dark) were recruited. Each was measured using 4 types of colour measuring methods (tele-spectroradiometer, spectrophotometer, digital camera and visual assessment) including 6 instruments and 2 sets of colour charts. The results from the former two types are summarised here. The results were analysed in terms of skin colour distribution, repeatability and inter-instrumental agreement between 4 skin groups in CIELAB coordinates and spectral domain. Introduction Skin colour has been one of the most extensively studies over the years. It has been involved with many applications such as photography, display, imaging, printing, medical, lighting, etc. It is important to reproduce skin colours to make them either preferable or accurate. So, many experiments were conducted using different colour measuring instruments. According to the target measured, they can be divided into non-contact and contact methods. For contact method, spectrophotometers are normally used, including a light source, a grating and a detector. They have the illumination/viewing geometry of either diffuse integrating sphere or 45o:0o. For measuring skin colours, caution should be taken to apply constant pressure to the surface because skin colour varies with different pressure applied [1]. Different from cosmetic, printing and imaging industries which are interested on the colour management and colour specification, the medicals are also interested in the pigments in the skin. Spectrophotometers are normally used to measure colours at a fixed skin location to ob tain the haemoglobin concentration [2,3] . The non-contact methods include tele-spectroradiometer (TSR), camera and visual assessment. They have been widely used for measuring skin colou rs. The visual assessment was also used by means of reference colours presented by a fan deck or a colour chart. A typical example is that De Rigal [4] measured skin colours to design skin colour charts as a visual aid for evaluating skin whitening product. It provides a fast and inexpensive method to evaluate the effect of clinic treatment or to find a cosmetic product to match the skin colour in stores. TSR measures the spectral power distribution (SPD) of a colour illuminated by a source, and spectrophotometer measures the spectral reflectance of a surface colour. More recently, colour calibrated digital cameras were used to measure colours of objects [5]. It can obtain the colour information of the whole region rathe r than a selected location like the other measuring methods. Although the above methods have been used, there is a lack of research to investigate their differences. Because of the importance of skin spectral database, ISO ISO/TR 16066-2003 Graphic Technology – Standard object colour spectra database for colour reproduction evaluation (SOCS) [6] provides a database including 51182 sets of spectral reflectance, for which 8213 of them are skin colours. There are 6 skin groups, provided by 5 organisations. Each subject was measured at forehead, cheek, neck, zygomatic region and arm. However, they did not define the instruments and their measuring conditions used. It can be found that the colour distribution covers a very large co lour gamut. So, it is not able to be used for further research. With the above in mind, the CIE has established a technical committee, TC 1-92 Skin Colour Database. It is aimed to investigate the uncertainty in skin colour measurement, to recommend protocols for good measurement practice, and to evaluate skin colour measurements that according with these protocols covering different ethnicity, gender, age and body location. This paper summarises the results of the two datasets, which carried out at Liverpool and Leeds Universities, UK, and Zhejiang University, China. The objectives are to report the performance of repeatability and inter-instrument agreement, to reveal the colour distributions for each instrument, and to compare the measuring results between two sites having same type of instrument, such as between two TSRs or between two SPs. It is hoped to provide a general understanding of colour variation using different instrument for measuring different skin groups. Experimental Data Two datasets accumulated at Liverpool/Leeds Universities and Zhejiang University, called UK data and China data, respectively. Table 1 shows the details of the data collected from each site. It includes number of subjects, male and female, of each skin group and the number of locations measured for each subject. Four measurement methods were used: TSR, spectrophotometer (SP), digital camera, and visual assessment. Only the measurement results from SP and TSR are investigated here, because these instruments were used in both experiments and are more widely used to accumulate skin colour database. Table 2 describes the specification about the 4 instruments used. Table 1 shows that there are 4 times more subjects participated in the UK experiment than that of China experiment. It can alsobe found that there is also a shortage of female subjects for the South Asian and Dark skin groups in both data. Table 1 The number of subjects in the UK and China experiments. UK data Chinese data Skin group Male Female Male Female