Skin colour data are important for many applications such as medical, imaging, cosmetics. The present study was aimed to collect a comprehensive skin colour database, and to study the impact of the skin colour on the variety of facial impression attributes. Although many researchers and engineers have collected skin data，few of them studied the skin colours to measure the same locations on a large number of subjects from different ethnic groups using the same colour measuring instruments. As for studying the impact of the skin colour on the visual perceptions, many studies investigated the impact of the skin colour on the attractiveness, health and youth. Limited previous studies investigated the impact of the skin colour on the other impression attributes.
The present study was divided into two experiments, Experiments 1 and 2. Experiment 1 was to accumulate the skin colour database, named the Leeds Liverpool skin colour (LLSC). It included skin colours of 188 people from four ethnic groups (Caucasian, Oriental, South Asian and African) and both genders. Three colour measuring methods were used to accumulate the skin colour of each subject’s 10 locations including facial locations (forehead, cheekbone, cheek, nose tip, chin and neck) and body locations (the back of the hand, inner forearm, outer forearm and fingertip). The colour measuring methods included a tele-spectroradiometer (TSR), a spectrophotometer (SP) and a set of skin colour chart used as a visual aid. Also, a characterised digital camera controlled by an imaging system was used to collect facial images. Before the data collection, the short-term repeatability of different settings of the TSR and the SP on measuring human skin colour in vivo was determined. And this was used to settle the measurement protocols of the two instruments. The LLSC database was later used to investigate the skin colour distribution between ethnic groups, between genders, between measuring methods. A skin whiteness and blackness scales based on the CIELAB L* and Cab* scales in CIELAB was developed by referencing the vividness and depth formulae, which was developed by Berns (2000). It was found that these scales and CIELAB hue angle can describe well the property of skin colour of each ethnic group.
Experiment 2 was to investigate the impact of the skin colour on the facial impression attributes. Based on the LLSC database, the gamut of skin colour was defined. Twenty-three attributes used to describe facial skin colours were accumulated. They were classified into two groups (appearance and impression). Two experiments were carried out on a monitor to understand the impact of the skin colour on the perceived facial impression attributes. The first experiment (Experiment 2.1) was to study the relationship between different attributes by 10 observers. The results showed that only four dimensions were required to describe skin facial colours, which were named Likeable, Sociable, Feminine and Youth. The health was also selected because the traditional Chinese medicine has interested in it. The second experiment (Experiment 2.2) was to scale facial images selected from two ethnic groups and both genders by using these five impression attributes by 24 Chinese observers. The experimental results showed that there were systematic patterns between the impression attributes and the whiteness and hue angle scales. There are some differences between these images for each impression. The ethnic group had an impact on the judgement, but the difference between the Oriental and Caucasian female images was limited. Finally, mathematical models were successfully developed to predict the impressions from the skin whiteness and hue angle data.
