摘要
人脸超分辨率重建是基于人脸图像的特性从低分辨率图像重建高分辨率图像的超分辨率问题。基于人脸图像的部件化构成特性,提出一种新的人脸超分辨率重建方法。该方法通过对非负矩阵分解进行典型相关分析获得更佳的全局人脸图像,通过高维耦合非负矩阵分解进行残差补偿获得更多细节信息。实验表明,该方法取得了令人满意的重建效果。
Face super-resolution is the specific super-resolution based on the characteristic of facial images,which reconstructs a high-resolution facial image from a low-resolution input.In view of the observation that facial images are made up of several relatively independent parts,novel face hallucination method is proposed,which can generate a better global face image by correlation-constrained non-negative matrix factorization(CCNMF),and preserve more high frequency details by High-dimensional Coupled Non-negative Matrix Factorisation(HCNMF).Experimental results verify the effectiveness of the method.
作者
李翊凡
郭常忠
Li Yifan;Guo Changzhong(HBIS Group Wusteel Company)
出处
《宽厚板》
2020年第2期46-48,共3页
Wide and Heavy Plate
关键词
超分辨率
非负矩阵分解
样本学习
Super resolution
Non-negative matrix factorization
Sample study
作者简介
李翊凡,男,2012年毕业于中国人民解放军信息工程大学信息技术应用与管理专业,工程师。
