基于小波变换的人脸表情特征提取

Wavelet-based Facial Expression Feature Extraction

首先通过直方图均衡化处理增强图像的整体对比度,使图像的细节更加清晰.通过离散余弦变换来降低图像特征维数、去除冗余信息、保留重要的低频信息.然后利用Gabor小波变换,选取不同的尺度和方向对人脸表情特征进行提取.最后通过实验结果对比证明预处理后的图片在进行小波变换时能节省大量的运算时间.

First,a histogram equalization process is used to enhance the overall image contrast to make image detail clearer.By discrete cosine transform we can reduce the image feature dimension,remove redundant information,retain an important low-frequency information.Then it uses the Gabor wavelet transform,selects a different scale and direction of facial expression feature extraction.Finally,by comparing experimental results,it proves that during the pre-image after wavelet transform we can save a lot of computing time.