非负特征基约束的人脸超分辨率

摘要

Principal Component Analysis (PC A) is commonly used for human face images representation in face super-resolution. But the features extracted by PCA are holistic and difficult to have semantic interpretation. In order to synthesize a better super-resolution face image with the results of the face images representation, we propose face a super-resolution algorithm with non-negative featrue basis constraint The algorithm uses the NMF to obtain non-negative featrue basis of face sample images, and the target image is regularized by Markov random fields, with maximum a posteriori probability approach. Finally, the steepest descent method is used to optimize non-negative featrue basis coefficient of high-resolution image. Experimental results show that, in the subjective and objective quality, the face super-resolution algorithm with non-negative feature basis constrait performs better than PCA-based algorithms.%主成分分析(PCA)是人脸超分辨率申常用的人脸图像表达方法,但是PCA方法的特征是整体的且难以语义解释.为了使表达的结果更好地用于合成超分辨率人脸图像,提出一种非负特征基约束的人脸超分辨率算法.该算法利用非负矩阵分解(NMF)获取样本人脸图像的非负特征基,结合最大后验概率的方法,对目标图像进行马尔可夫随机场正则约束,最速下降法优化得到高分辨率人脸图像的非负特征基系数.实验结果表明,在主客观质量上,非负特征基约束的人脸超分辨率算法的性能胜过基于PCA的算法.