多级FFD融合超分辨率重建的视频人脸识别

针对视频人脸识别中由于人脸畸变、表情变化等非刚性变化导致无法精确配准和重建的问题,提出一种基于多级自由变形配准的超分辨率重建算法。首先,利用低分辨率FFD网格全局配准,引入边缘配准度量到差平方总和准则;然后,将全局配准后的图像和基准图像划分成一系列对应子图对,使用高分辨率FFD网格对相关系数小的子图对进行局部配准;最后,采用凸集投影算法对多帧低分辨率图像重建SR人脸图像,并利用支持向量机分类器完成人脸识别。在标准视频库Choke Point和自己搜集的人脸视频库上的实验结果表明,在人脸畸变和表情变化很大的情况下,本文算法也能够精确配准和重建人脸图像,相比其它几种视频人脸识别算法,本文算法取得了更好的识别效果。

The Research of Video Face Recognition Using Super-resolution Reconstruction Based on Multi-level FFD Registration

The non-rigid change of deformed face and expression changes greatly affect the accuracy of registration and reconstruction in video face recognition, for which super-resolution reconstruction algorithm based on multi-level free form deformation registration, is proposed. Firstly, low-resolution FFD grid is used for global registration, edge registration measure is applied into the sum of squared difference criterion to emphasize the contribution of edge infor_mation for registration. Then, the global registration image and reference image is divided into a series of corresponding sub-image pairs and calculate the correlation coefficient of each pair and high-resolution FFD grid is used to local reg_ister the small value correlation coefficient sub-image pairs. In the registration process of optimization. Finally, the al_gorithm of project onto convex sets is used to reconstruct SR face image through several lower solution image se_quences, and support vector machines classifier is used to finish face recognition. Experimental results on standard video database ChokePoint and a database searched by ourselves show that proposed algorithm can register and recon_struct face image accurately in the condition of great face deformation and expression change, it has better recognition efficiency than several other video face recognition algorithms.