基于差分和肤色图像的人脸检测算法

提出并实现了一种基于差分和肤色图像的人脸检测算法。该算法利用视频图像的运动信息,在帧间差分的基础上进行二值图像边缘提取,确定目标在原图像中坐标位置,然后设计肤色模型表征人脸颜色,采用彩色图像的色系坐标变换进行人脸的准确定位。该算法的优点是可将运动信息序列图像中与人脸肤色相似的固定区域删除,在目标跟踪和运动检测上,不仅能有效地抑止背景噪声,减少误检率,而且还能缩小人脸检测范围,加快检测速度。实验表明,该算法可行、有效。

Facial Feature Locating Algorithm Based on Inter-Frame-Differential and Skin Color-Space Segmentation

In this paper, we proposed a facial feature locating algorithm based on inter-frame differential and skin color-space segmentation, First, we utilize the moving information to calculate the binary inter-frame differential image and extract edge and locate a gross position of the target. Secondly, we design the skin color model and use the color space transform to accurately locate human-face, using inter-frame-differential algorithm can remove possible disturbing background in complicate environment which is quite possible in application, Using skin-color can further restrict the interest region to face. Doing feature extraction after the two-step preprocessing could not only remove environmental noise and decrease false identification, but also, as it largely down-size the searching-area, it actually increased the total processing speed. The results of the applications show that based on inter-frame-differential and skin color-space algorithm is practical and efficient.