高分辨率预期图像与低分辨率输入图像分别位于高斯金字塔的第0层和第i层.超分辨率过程据此被分成i-1个子过程,逐层进行,即将第k层低分辨率图像的超分辨率估计图像(位于第k-1层)作为第k-2层高分辨率预期图像的输入,直至产生第0层高分辨率估计图像为止.在每个超分辨率子过程中,采用基于patch的搜索策略.文中对这种基于金字塔策略的人脸超分辨率方法进行了实验,结果表明该方法能够达到很好的超分辨率效果.
[1] FREEMAN W, PASZTOR E. Learning low-level vision [A]. Kerkyra, Corfu, Greece 1999.
[2] BAKER S, KANADE T. Limits on super-resolution and how to break them [A]. Hilton Head, SC, USA 2000.
[3] LIU CE, SHUM HEUNG-YEUNG, ZHANG CHANG-SHUI. A Two-Step Approach to Hallucinating Faces:Global Parametric Model and Local Non-Parametric Model [A]. Vancouver, Canada 2001.
[4] Yao LU, Minoru INAMURA. Super-Resolution Image Pyramid [J]. IEICE transactions on information and systems 2003, 8(8).
[5] WANG Qiang, TANG Xiaoou, SHUM H. Patch based blind image super resolution [A]. 北京 2005.
[6] PHILIPS P J, MOON H, RAUSS P. The FERET evaluation methodology for face-recognition algorithms [A]. San Juan, Puerto Rico 1997.
