摘要： 提出一种基于直方图统计学习的人脸检测方法，对人脸样本和非人脸样本进行小波变换，运用一组小波系数来表征各种人脸特征信息。统计每个训练样本的直方图分布，用于描述人脸和非人脸外观特征的概率分布，每个直方图表示一组小波系数与它们在人脸中位置的联合概率密度。该方法可以准确检测自然场景中的多幅人脸，对侧面人脸有很好的检测效果。

关键词: 人脸检测, 小波变换, 直方图统计, 贝叶斯决策规则, Adaboost算法

Abstract: This paper presents a face detection method based on histogram statistical learning. It does wavelet transform to the face samples and non-face samples and uses groups of wavelet coefficients to represent all kinds of attributes of face. The probability distribution of visual attributes of face and non-face is represented by statistics of distribution of histograms for each training sample, and each histogram represents the joint probability distribution of a subset of wavelet coefficients and their position on the face. This method can detect multiple faces in the natural scenes accurately. It gives a good detection performance for profile-view face detection.

Key words: face detection, wavelet transform, histogram statistical, Bayes decision rule, Adaboost algorithm

中图分类号: 

• TP391.41