基于小波变换的人脸图像压缩

传统的人脸识别方法,在对压缩的人脸图像进行识别时,要先进行图像的“全解压”操作,然后再进行识别,这无疑大大限制了识别系统性能的提高.为提高压缩人脸识别中压缩域特征提取的有效性和识别率,提出一种支持在压缩数据中提取识别特征的人脸图像压缩算法,反对称双正交小波变换子带编码的方法.与DCT法相比较,在获得高压缩比的同时,较好地保留了人脸图像的识别特征和主观质量.

Compressed Facial Image based on Wavelet Transform

The traditional methods of face recognition was “complete decompression” operation for the image first when the face image on comp ressed domain was recognize, which must limit identification system to enhance performance greatly. For improving the effectiveness and identifying rate of feat ure extraction in the face image on compression-domain, according algorithm for facial image was presented in this work, which supports feature extraction in compression-domain, the method of anti-symmetric biorthogonal wavelet. This method gained the high compression ratio. At the same time, the face image identifica tion characteristics and subjective quality was reserve preferably in comparison with DCT.