Multicolor Skin Modeling with Application to Skin Detection

In literature, different researchers consistently argue different color models to be the best choice for skin detection. However, to the best our knowledge, no significant work has been reported in the literature that attempted to utilize more than one color model for skin detection and evaluate the performance for identifying adult image contents. In this paper, we propose a rapid statistical framework for skin detection with an application of adult image identification, based on Multicolor Skin Modeling (MSM). From a high level, the proposed approach proceeds in two consecutive steps (levels). At the first step, the underlying goal is to identify and isolate skin regions of interest (ROI) in each image. At the second step, the suspected skin regions are fed into a specialized statistical analyzer which first extracts some statistical features form these regions. Then, an SVM classifier is used on the extracted features to verify the presence/absence of an adult content in images. Quantitative evaluation shows that our method compares favorably with the state of-the-art methods in terms of detection rate and false alarm, while reducing the computational costs by at least a factor of six compared with Forsyth’s method.