Detection of faces in a color natural scene using skin color classification and template matching

In this paper, an efficient algorithm for the detection of human faces in a color nature scene is proposed. The proposed algorithm consists of two sections. The first section is the segmentation of skin color regions using color information. Large data of chrominance like skin pixels is collected and analyzed in a training process to study its distributions. Then a binary color map is obtained by applying the skin/non-skin color classification algorithm. Thereafter, we use a boost filter algorithm to generate the facial features (eyes, mouth), then are merged with the obtained color map. The second section is to perform face verification by aligning a model of faces representing a front view face in a template-matching process. The fitness is calculated between the model and the candidate face regions to verify the elliptical shape of face regions. 50 images are used to examine the algorithm performance, and experimental results demonstrate that our algorithm can deal with different sizes, and different lighting conditions problems. A comparison between our algorithm and a well known related algorithm is demonstrated. The experimental results reveal that the proposed algorithm gives better results than the traditional algorithm in terms of accuracy, and time consumption