基于肤色模型的人脸检测算法研究

利用一种基于肤色的2D Gaussian模型和一定的先验知识实现了人脸快速检测。首先对一幅图像进行去噪、光线补偿等预处理,然后根据颜色空间特性建立适宜的肤色模型,检测出大致的肤色轮廓,对面部特征根据先验知识实现定位,从而剔除非脸部区域,最后对该算法进行实验验证。实验表明,该算法可以实现人脸的快速检测,降低了误报率,具有一定的优越性。

An Algorithm Research on Face Detection Based on Skin Model

Abstract: Fast face detection, based on skin 2D Gaussian model and certain prior knowledge, is realized in this paper. First, the image is preprocessed such as removing noise and light compensating. Secondly, a suitable skin model is built based on the color space, in which the skin color contour is detected out roughly and facial features are located according to the prior knowledge so that other areas are eliminated. Last, the algorithm is verified through experiments. Results show that it can realize the fast face detection, reduce the rate of false alarm and have certain superiority to other algorithms.