Face detection by polling

There are some aspects considered in this work; (1) human skin colors are a special category of colors, distinctive from the colors of most other natural objects; (2) this category has been found as a cluster in different color spaces; (3) the skin color variations between people are mostly due to differences in intensity; and (4) face detection based on skin color is a faster method as compared to other face detection techniques. In this work, we detect faces by carrying out skin color detection in four different color spaces: HSI, YCbCr, YES and RGB. Once the skin color clusters have been detected in every color space, we use some cluster characteristics such as size and position to preserve the regions where the face is possible positioned. Finally, we employ a polling strategy with regions preserved in all color spaces to determine the final region where the face is detected and located. The system has been tested with images and video and it has not failed