Demographic Classification Using Skin RGB Albedo Image Analysis

Wei Chen, Miguel Viana, Mohsen Ardabilian, Abdel-Malek Zine

Abstract: Age, gender and skin type classification of demographics using common imaging techniques is costly and does not provide good performance. We propose an approach based on skin RGB albedo image analysis for demographic classification. The diffuse albedo uses inherent skin properties which prevail over illumination conditions variation despite being based on visual perception. The method was tested using skin samples from multiple facial regions to evaluate their performance for classification. Moreover, the application of a fusion algorithm using albedo data from each of the facial regions improved the overall performance resulting in rates above 90% accuracy in age, gender and skin type categories.

External IDs:dblp:conf/iciap/ChenVAZ17