|
Traditional methods of human skin detection are usually based on an RGB camera. To handle the problem of metameric color, multispectral images have attracted more attention. In this paper, a multispectral imaging system is developed for skin detection. In order to simplify the system, a band selection algorithm has been used to choose the effective bands from 31 bands within 400nm to 1000nm. In the test for discriminating skin and skin-like objects, two bands around 800nm and 420nm are selected. A significantly better detection performance has been obtained compared with the a simulated three-band RGB images.Two main contributions in this paper are presented. First, as we know, it is the first time that a band selection algorithm has been used for skin detection. Second, by using a conditional probability scheme, the band selection method can reduce the computational complexity caused by the calculation of high-dimensional distance measures. |
|
Keywords:Skin Detection; Band Selection; Multispectral |
|