A robust method for face recognition based on Android platform is proposed.Ac cording to the variability of scene,firstly face image normalization processing through eye detection,pupil orientation, GIC correction and histogram equalization algorithm, and these finally give to a satisfactory practical effect. Then, it extracts the LBP global histogram features of a face image,then subdivides the whole face image into a number of blocks of the same size and extracts the LBP local histogram features of each block finally combines the global features with the local histogram in specified order as the total feature of face image. The proposed approach was tested using ORL face database based on Androidplatform.Experimental results demonstrate that the LBP feature has an encouraging performance for face recognition.%提出一种基于Android平台实现的具有较强鲁棒性的人脸识别方法.根据终端用户拍摄场景的多变性,首先通过人眼检测、瞳孔定位、GIC校正和直方图均衡化算法对人脸图像进行归一化处理,达到比较好的实用效果 ;其次利用局部二值模式(LBP)提取全局直方图特征,再将图像划分为若干大小相同的子区域,提取每个子区域的LBP直方图,最后将全局和局部直方图按一定的顺序结合作为人脸图像的最终特征.在Android平台上用ORL人脸数据库对该方法进行测试,结果表明具有较好的识别性能.
展开▼
