|
|||||
|
|
| 图像超分辨率重建中的运动模型选择 | |
| 引用本文： | 张雪松,江静,彭思龙.图像超分辨率重建中的运动模型选择[J].华北科技学院学报,2009,6(1):58-61,65. |
| 作者姓名： | 张雪松 江静 彭思龙 |
| 作者单位： | 1. 光电系统信息控制技术国家级重点实验室,河北,三河,065201;中国科学院自动化研究所,北京,100090 2. 华北科技学院机电系,北京,东燕郊,101601;中国矿业大学(北京)北京,100083 3. 中国科学院自动化研究所,北京,100090 |
| 基金项目： | 国家科技支撑计划重点项目 |
| 摘 要： | 愕统的超分辨率方法通常假定一个低分辨率图像序列所服从的固定的运动模型，怛是这种假设并不能覆盖实际应用中出现的各种运动情况，从而严重影响图像的重建质量。本文提出一个“虚拟像机”的概念，将序列中的每一幅低分辨率图像映射到一张高分辨率参考图像的网格上，并引入运动模型选择来同时估计出两视点几何关系和相关参数。文中给出了应用GRIC（Geometric Robust Information Criterion）准则确定两视点几何关系的方法。 |
| 关 键 词： | 模型选择 运动估计 超分辨率 |
On Motion Model Selection in Image Super-resolution Reconstruction |
|
| ZHANG Xuesong,JIANG Jing,PENG Silong.On Motion Model Selection in Image Super-resolution Reconstruction[J].Journal of North China Institute of Science and Technology,2009,6(1):58-61,65. | |
| Authors: | ZHANG Xuesong JIANG Jing PENG Silong |
| Institution: | 1. National Laboratory of Electro-Optics System Technology,Sanhe Hebei 065201; 2. Institute of Automation, China Academy of Sciences,Beijing 100090;3. Department of Mechanies and Electricity Engineering North China Institute of Science and Teehnology, Yanjiao Beijing-East 101601 ; 4. China University of Mining Technology (Beijing) ,Beijing 100083) |
| Abstract: | Conventional SR algorithms usually assume that a fixed motion model is followed by all LR image sequences,which is,however,an impractical assumption for many real applications,and therefore affects the reconstruction quality heavily.This paper proposes a concept,Virtual Camera,to map each LR image onto a high-resolution reference image grid and introduces motion model selection to simultaneously estimate the epipolar geometry and the relevant parameters.The method using the GRIC(Geometric Robust Information... |
| Keywords: | Model Selection Motion Estimation Super-resolution |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ | |