Human Skin Detection using Combination of Color Spaces in Color Images

Enhancing the performance of human face detection, tracking and recognition has been done by skin detection considering it as the fundamental step for the several decades. Of the evolved skin detection techniques, skin cluster classifier based on the human skin tone color are considered along with the concept of combination of two color spaces. This paper presents a novel method of combining more than one color space for human skin segmentation based on the piecewise linear decision boundary classifier and provides an analogous study on the use of color spaces for the segmentation of human skin regions under inconsistent illumination conditions, basically altering skin tone color and varying lighting conditions. The color spaces that are used for this study are: HSV, HSL, LMS, LSLM, XYZ, XYEz, YCC, YPbPr, and YDbDr. Combinations of two color spaces are used for the analysis of success rate for the human skin detection region. Finally, the comparison is made between single and combinations of color spaces to find out the best possible color space(s) either single or combination of two color spaces that has given clear discrimination between skin and non-skin pixels in color image. The successfulness of the color space(s) or combination of color space(s) depends on the false positive and false negative rate obtained after the extensive experimental evaluation over the commonly used sample face database.