现代电子技术2017，Vol.40Issue(13)：57-61,5.DOI:10.16652/j.issn.1004-373x.2017.13.015
基于在线字典学习的人脸超分辨率重建
Human face super-resolution reconstruction based on online dictionary learning
摘要
Abstract
Aiming at the problems of more noisy points and artifacts,and poor noise robustness existing in the learning-based human face super-resolution algorithm,a human face super-resolution reconstruction algorithm based on online dictionary learning is proposed. The human face image set is taken as the training library. The online dictionary learning method is used to improve the accuracy of dictionary training. The regularization parameter λt of the dictionary learning phase is regulated indepen-dently,and regularization parameter λr in the reconstruction stage of the sparse coefficients is solved to get the optimal overcom-plete dictionary and sparse coefficients for image reconstruction. The experimental results show that the peak signal-to-noise ratio (PSNR) of the target image of the proposed algorithm is 0.85 dB higher and the structural similarity is 0.0133 higher than that of the same type sparse coding super-resolution algorithm averagely,which can restrain the noisy point and artifact effectively. The application result of noisy human face image shows that the PSNR is decreased smoothly when the noise level is increased, which can improve the robustness against noise while promoting the performance of face super-resolution.关键词
在线字典学习/超分辨率重建/含噪人脸图像/稀疏编码Key words
online dictionary learning/super-resolution reconstruction/noisy human face image/sparse coding分类
信息技术与安全科学引用本文复制引用
刘芳华,阮若林,王建峰,倪浩..基于在线字典学习的人脸超分辨率重建[J].现代电子技术,2017,40(13):57-61,5.基金项目
国家自然科学基金项目(61271256) （61271256）
湖北省自然科学基金项目(2015CFB452) （2015CFB452）
湖北省高等学校优秀中青年科技创新团队计划项目(T201513) （T201513）
湖北省教育厅科研计划指导性项目(B2015080) （B2015080）
