Abstract—A new skin-based color space is introduced in this paper that can be effectively used for skin detection. Converting to the new space is performed through a linear matrix multiplication. The mission is to address how to obtain the weights of the conversion matrix so that the new space can discriminate skin regions. Quadratic programming is used to obtain the weights of the conversion matrix by optimizing the distances between the color components of some sample skin and non-skin pixels. The new Quadratic Programming-based Color space (QPC) uses a new form of distance metric under some specific criteria. Otsu's thresholding method as a post-processing technique is applied to the converted image in QPC to binarize the image for detection purposes. Common skin image benchmarks is used to evaluate the performance of the proposed method. Experiments shows that the proposed method hands in a conversion matrix that can be used effectively to discriminate skin in a wide range of images, and produces better results in comparison to some standard color spaces and classification algorithms.

Index Terms—New color space, skin detection, quadratic programming, Otsu's thresholding, connected component analysis.

Ali Khayati is with Department of Computer & IT Engineering, Parand Branch, Islamic Azad University, Tehran, Iran.
Saeed Mirghasemi is with School of Engineering and Computer Science, Victoria University of Wellington, Wellington, New Zealand.

Cite:Ali Khayati, Saeed Mirghasemi, "A New Color Space for Color Skin Detection Using Quadratic Programming," International Journal of Computer and Electrical Engineering vol. 10, no. 1, pp. 31-45, 2018.