摘要： 基于多角度人脸图像的性别分类是计算机视觉领域的一项具有挑战性的研究课题。为了提高多角度人脸性别分类的准确率，提出一种新的局部Gabor二值映射模式特征提取方法。该方法结合了局部二值模式、图像空间信息以及Gabor小波变换的幅值信息，对图像噪声、光照变化和人脸角度变化均具有一定的鲁棒性。在中科院CAS-PEAL人脸数据库上进行的实验表明，在所有9种不同角度的人脸图像中，该方法取得了95%的最高平均准确率。

关键词: Gabor滤波器, 二值映射模式, 性别分类, 支持向量机

Abstract: Multi-view gender classification, which is based on facial images, is one of the most challenging problems for computer vision researchers. This paper proposes a novel approach, Local Gabor Binary Mapping Pattern(LGBMP), to improve the correct classification rate for multi-view gender classification. The proposed approach which combines local binary pattern histogram, spatial information and the magnitude part of Gabor filter is robust to noise and local image transformations caused by variations of illumination and pose. Experimental results on the CAS-PEAL face database show that the proposed LGBMP achieves the highest correct classification rate of 95% on all of the 9 face poses.

Key words: Gabor filter, binary mapping pattern, gender classification, Support Vector Machine(SVM)

中图分类号: 

• TP391.4