为了降低人脸Gabor特征的维数,提出了一种新的基于Gabor幅值的纹理表征(GMTR)方法用于人脸识别.GMTR由伽玛分布(TD)拟合Gabor幅值的分布来刻画,拟合的TD参数作为纹理特征.在FERET和Yale人脸库上的实验结果显示GMTR的识别性能优于传统的Gabor幅值特征,表明纹理特征具有更强的鉴别力.%To reduce the dimensionality of the Gabor features, this paper presented a novel approach called GMTR for face recognition. Characterized GMTR by using ΓD to model the Gabor magnitude distribution. The estimated model parameters served as texture representation. Experimental results performed on FERET and Yale databases show that GMTR is superior to traditional Gabor features in terms of recognition accuracy, which demonstrates the Gabor texture features have more discriminative power.
展开▼