| 基于聚类学习的超分辨率方法 |
| |
| 引用本文： | 吉高云,韩华.基于聚类学习的超分辨率方法[J].山东交通学院学报,2007,15(2):82-86. |
| |
| 作者姓名： | 吉高云 韩华 |
| |
| 作者单位： | 1. 山东交通职业学院管理系,山东 潍坊,261206
2. 中国科学院自动化研究所国家专用集成电路设计工程技术研究中心,北京,100080 |
| |
| 摘 要： | 在高分辨率空间对图像局部结构进行k均值聚类，估计最小二乘意义下的插值滤波器参数，然后根据估计参数来增强图像的分辨率。同时借鉴区域连接增长的思想解决分类歧义问题。计算结果显示，该方法优于传统的基于学习的方法，图像边缘清晰、细节恢复多。
|
| 关 键 词： | 图像超分辨率 学习 聚类 插值滤波器 相关度量 区域连接增长 |
| 文章编号： | 1672-0032（2007）02-0082-05 |
| 修稿时间： | 2006-05-16 |
A Method of Super Resolving Power Based on Clustering Study |
| |
| JI Gao-Yun,HAN Hua.A Method of Super Resolving Power Based on Clustering Study[J].JOURNAL OF SHANDONG JIAOTONG UNIVERSITY,2007,15(2):82-86. |
| |
| Authors: | JI Gao-Yun HAN Hua |
| |
| Institution: | 1. Department of Management, Shandong Jinotong Vocational College, Weifang 261206, China;2. Research Center of National Special-purpose Integrate Circuit Design Engineering Technology, Automatization Research Institute of China Academy of Sciences, Belting 100080, China |
| |
| Abstract: | To cluster K equal value for image partial structures in high resolving power space, the parameters of inserting-value filters under the minimal quadratic meaning are evaluated. Then resolving power of images is heightened according to the evaluated parameters. Meanwhile the problem of classifying ambiguity is solved using the regional connection increase for reference. Calculating results show that this method, with images that have clear-cut edges and many details that can be resumed, is superior to the traditional methods based on study. |
| |
| Keywords: | images with super resolving power study clustering inserting-value filter relevant measurement regional connection increase |
| 本文献已被 维普 万方数据 等数据库收录！ |
