A discriminative fusion framework for skin detection

Skin detection is one of the preprocessing steps of machine vision applications. In this paper, a discriminative fusion framework is proposed for skin/non-skin classification of image pixels. The method utilizes conditional random fields (CRFs) to statistically combine the information of original raw image with the decisions made by a group of intermediate detectors to improve the accuracy and robustness of the detection task. The experimental result shows the success of the proposed fusion approach in comparison to the primary detectors.