An Adaptive Detection Method of Multiple Faces

The appearance of multiple faces is influenced by abnormal exposure, interfering backgrounds or fake objects greatly in the color face image. A multiple-face detection method based on the adaptive dual skin model and improved fuzzy C-mean clustering was presented in this study. First an adaptive skin-color model and an adaptive skin-probability model were built to acquire the skin likelihood for clustering, the adaptive initial clustering centers, and the adaptive clustering weights. Then the skin-likelihood image was segmented dynamically by improved fuzzy C-mean clustering. Finally the multiple-face targets were distinguished and extracted by jointly using the effective areas, circumferences and circularities of connected targets. Experiment showed that the algorithm had good results and high speed, accuracy, and adaptability of face detection.