计算机应用 ›› 2016, Vol. 36 ›› Issue (6): 1654-1658.DOI: 10.11772/j.issn.1001-9081.2016.06.1654

• 虚拟现实与数字媒体 • 上一篇    下一篇

基于局部回归模型的图像超分辨率重建

李欣, 崔子冠, 孙林慧, 朱秀昌   

  1. 南京邮电大学 通信与信息工程学院, 南京 210003
  • 收稿日期:2015-11-30 修回日期:2016-01-14 出版日期:2016-06-10 发布日期:2016-06-08
  • 通讯作者: 李欣
  • 作者简介:李欣(1981-),女,安徽巢湖人,讲师,博士,主要研究方向:图像超分辨率重建、多媒体通信;崔子冠(1982-),男,河南郑州人,副教授,博士,主要研究方向:视频编码与传输、图像处理;孙林慧(1979-),女,山西临汾人,副教授,博士,主要研究方向:信号处理、现代信号通信;朱秀昌(1947-),男,江苏丹徒人,教授,博士生导师,硕士,主要研究方向:图像处理、多媒体通信。
  • 基金资助:
    国家自然科学基金青年基金资助项目(61501260);江苏省自然科学基金资助项目(BK20130867,BK20140891);江苏省高校自然科学基金资助项目(13KJB510020);江苏省普通高校研究生科研创新计划项目(CXLX11_0406,CXLX12_0474)。

Image super-resolution reconstruction based on local regression model

LI Xin, CUI Ziguan, SUN Linhui, ZHU Xiuchang   

  1. College of Telecommunications & Information Engineering, Nanjing University of Posts and Telecommunications, Nanjing Jiangsu 210003, China
  • Received:2015-11-30 Revised:2016-01-14 Online:2016-06-10 Published:2016-06-08
  • Supported by:
    This work is partially supported by National Natural Science Fund for Distinguished Young Scholars (61501260), the Natural Science Foundation of Jiangsu Province (BK20130867, BK20140891), the Natural Science Foundation of the Jiangsu Higher Education Institutions of China (13KJB510020), the Postgraduate Innovation Program of Scientific Research of Jiangsu Province (CXLX11_0406, CXLX12_0474).

摘要: 针对基于稀疏重建的图像超分辨率(SR)算法一般需要外部训练样本,重建质量取决于待重建图像与训练样本的相似度的问题,提出一种基于局部回归模型的图像超分辨率重建算法。利用局部图像结构会在不同的图像尺度对应位置重复出现的事实,建立从低到高分辨率图像块的非线性映射函数一阶近似模型用于超分辨率重建。其中,非线性映射函数的先验模型是直接对输入图像及其低频带图像的对应位样本块对通过字典学习的方法得到。重建图像块时利用图像中的非局部自相似性,对多个非局部自相似块分别应用一阶回归模型,加权综合得到高分辨率图像块。实验结果表明,该算法重建的图像与同样利用图像具有自相似性的相关超分辨率算法相比,峰值信噪比(PSNR)平均提高0.3~1.1 dB,主观重建效果亦有明显提高。

关键词: 超分辨率, 局部回归, 字典学习, 稀疏重建, 非局部自相似

Abstract: Image Super-Resolution (SR) algorithms based on sparse reconstruction generally require external training samples. The shortcoming of these algorithms is that the reconstruction quality depends on the similarity between the image to be reconstructed and the training sample. In order to solve the problem, an image super-resolution reconstruction algorithm based on local regression model was proposed. Using the fact that the local image structure would repeat in the corresponding position of different image scales, a first-order approximation model of the nonlinear mapping function from low to high resolution image patches was built for super-resolution reconstruction. The prior model of the nonlinear mapping function was established by handling the in-place example pair of the input image and its low frequency band image with dictionary learning. During the reconstruction of the image block, the non-local self-similarity of image was used and the first-order regression model was applied to multiple non-local self-similarity patches respectively, the high-resolution image patch could be obtained through weighted summing. The experimental results show that, compared with other super-resolution algorithms which also make use of image self-similarity, the average Peak Signal-to-Noise Ratio (PSNR) of the reconstructed images of the proposed algorithm is increased by 0.3~1.1 dB, and the subjective reconstruction effect of the proposed algorithm is improved significantly as well.

Key words: Super-Resolution (SR), local regression, dictionary learning, sparse reconstruction, non-local self-similarity

中图分类号: