Skin Detection Using a Statistical Color Spaces Fusion Model

Skin detection is the preliminary stage of manycomputer vision applications. In this paper, a statistical fusionmodel is proposed for detecting skin regions in arbitrary images.We used conditional random field (CRF) to statistically combinethe information of different color spaces and model the spatialrelationship between image pixels. The conditional probabilitydistribution of labels (skin or non-skin) given differentrepresentations of raw images in various color spaces is modeled.For a test image, the label image is inferred using the learnedCRF model. The experimental results verify the success of ourproposed method.

کلیدواژه ها:

Skin pixel detection ، Color spaces ، Conditional random fields (CRFs)