A face detection solution integrating automatic and user assisted tools

Abstract: Knowledge about the number, position and size of human faces present in a given image or video is valuable information for their content-based description. This paper proposes a two-stages solution for the face detection problem using both colour and shape information. The first stage consists in a pixel-based colour segmentation algorithm of the HSV colour space. For this purpose, a study about the skin-colour distribution of faces in natural images is first presented. The second stage recognises that an ellipse can approximate the outline of a human face shape in several positions. In this stage, an algorithm based on the Hough Transform is used to perform ellipse detection with some shape constraints. Since the automatic detection tool does not perform satisfactorily all the time, this tool is complemented with a user-assisted tool allowing the user to improve the automatic detection results. The system implemented has been tested for a wide variety of natural sequences and the experimental results show that the algorithm is robust to changes of position, size, orientation and number of faces in an image