基于多尺度等价模式LBP的人脸表情识别

Facial Expression Recognition Based on Multi-scale Equivalent LBP

人脸表情识别是近年来计算机视觉和模式识别领域的一个研究热点,论文提出了基于多尺度等价模式LBP(Local Binary Patterns)的人脸表情识别方法。首先,为了消除光照影响,对图像分别进行对数变换和直方图均衡化处理,并按照一定的比例关系融合。然后,提取图像的等价模式LBP谱图,对图像进行多尺度分解。为了减少运算量,提高识别率,仅以人眼和嘴部区域的LBP直方图作为表情特征向量。最后,对重点区域的特征向量进行了加权处理,以加权欧氏距离对表情进行分类。论文用JAFFE表情库进行了测试实验,实验结果表明,该方法识别率较高,速度快。

Facial expression recognition is a hot research topic in the field of computer vision and pattern recognition in recent years.In this paper a novel method based on multi-scale equivalent local binary patterns for facial expression recognition is proposed.First of all,in order to eliminate light effects,the logarithmic transformation and histogram equalization processing are used respectively on images,and two images preprocessed are fused according to certain proportion relationship.Then,the equivalent model of LBP spectrogram is extracted and the image is divided in multi level.For reducing the computational complexity and improving the recognition rate,only the LBP histograms of eye and mouth areas are reserved as a facial expression feature vector.Finally,the feature vectors of the key area are weighted and Euclidean distance is used for expression classification.This paper introduced the test experiment with JAFFE facial expression library,the experimental results show that the method proposed in this paper improve recognition rate and the operation speed is faster.