基于边缘方向直方图的Adaboost人脸检测

在传统的Adaboost分类器算法中,采用色彩直方图或空间直方图作为提取特征,但这并不能充分描述图像目标的特征,导致了目标检测和跟踪出现偏差。提出了一种基于边缘方向直方图的Adaboost人脸检测算法,使用空间分布和纹理信息作为提取特征。实验结果表明,该方法与传统Adaboost方法相比,准确率明显提高,而速度相当。

Adaboost for Face Detection Based on Edge Orientation Histograms

Color histogram and spatial histogram are widely used as characters in traditional Adaboost classifiers,however,they are not sufficient for describing the information of the objects.This causes error in object detection and tracking based on Adaboost.Presents a new Adaboost algorithm based on edge orientation histograms,using space distributing and texture information as characters,and is used for face detection.Experiments proved the new algorithm is more accuracy than and as fast as tradition ones.