Face Detection Based On Salient Region Extraction

Face detection (FD) is always the research hot point in the area of computer vision, and is the first step for unconstrained face recognition system. Given an arbitrary image, the main concern of face detection is to determine whether or not there are any faces in the image and, if present, return the image location and extent of each face. Currently, research on face detection has received considerable achievements, but some critical problems in this field are still needed to be resolved. Two novel face detection algorithms based on salient region extraction are presented in this paper.Firstly, visual attention model and salient region model are introduced in this paper, and a comparison has been made between two different algorithms of feature selection and feature area extraction. By combining three salient features with Adboost algorithm, we get the salient region of face in the complex background. The method overcomes the drawbacks existing in the original rules to generate saliency map and improves the color salient feature. Then the accurate face detection is carried out by use of SNoW classifier. In several experiments, the parameters of our model are the best in most cases, and the presented method for face detection can achieve the better consequences in the complex background and multiple-face images.Secondly, this paper introduces the basic theory and characteristics of neural network (NN). According to the special particularities of FD, it describes how to design an effective NN architecture with some theory analyzing and a series of experiments. In addition, a new merging solution is presented to remove duplicate detections. After the salient regions of test image are extracted by three salient features, neural network is used to detect faces. Comparing the presented method and the NN face detection based on skin color area and salient area, the experiments indicate the effect of the former is better than the latter obviously.