期刊文献+

基于双向光流估计的高动态范围图像去模糊

High dynamic range image deblurring with bidirectional optical flow estimation
下载PDF
导出
摘要 普通RGB图像是单曝光的低动态范围图像,无法呈现场景中的整个动态范围,因为在高曝光图像中存在过曝光区域而低曝光图像中存在欠曝光区域。高动态范围图像将同一场景不同曝光程度的低动态范围图像进行融合,以此弥补图像中存在过曝光和欠曝光区域,使图像细节信息更加丰富。但在高动态范围图像拍摄过程中,由于相机抖动或者快速运动物体曝光时间太短导致低动态图像模糊,进而导致合成的高动态范围图像也模糊。针对此问题,考虑到高动态范围图像由多帧除去运动物体以外背景相同的不同曝光程度的低动态范围图像合成,本文提出基于直方图匹配的双向光流方法实现高动态范围图像去模糊,结合多帧低动态图像中运动物体的时序信息构建运动光流能量方程,求解清晰图像,最终合成清晰高动态范围图像。实验表明通过此方法能实现高动态范围图像的去模糊,且与其余的单帧图像和多帧图像去模糊方法相比,本方法的效果求得去模糊的图像更加清晰。 The RGB images with a single exposure are often of low dynamic range so that the entire dy-namic range in the scene cannot be well presented.Commonly,there are over-exposed areas in high-exposure images and under-exposed areas in low-exposure images.High dynamic range images can be formed by fusing several low dynamic range images with different exposure in the same scene to compensate for overexposed and underexposed regions so that abundant details can be preserved.However,when shooting for high dynamic range images,camera shaking and quickly moving objects with short exposure time often result in blur in synthesized images.Thus,de-blurring is an important problem to deal with.Considering that a high dynamic range image is synthesized by a series of low dynamic range images with different exposure degrees of similar background,we propose to use a bidirectional optical flow method combining with histogram equalization for high dynamic range image de-blurring.The energy flow equation to optimize the solution of the latent image is constructed based on the motion optical flow.Experiments show that this method achieves clear high dynamic range images of good image quality.In comparison,our method performs better than other de-blurring methods using multiple continuous image frames or a single image.
作者 徐倩 钱沄涛 XU Qian;QIAN Yuntao(College of Computer Science and Technology,Zhejiang University,Hangzhou 310027,China)
出处 《中国体视学与图像分析》 2019年第4期342-351,共10页 Chinese Journal of Stereology and Image Analysis
基金 科技部科技创新2030新一代人工智能重大项目(2018AAA0100500)。
关键词 高动态图像 去模糊 光流 high dynamic image deblurring optical flow estimation
  • 相关文献

参考文献2

二级参考文献18

  • 1Fergus R, Barun S. Removing camera shake from a single photograph [J].ACM Transactions on Graphics (TOG), 2006,3(25) : 787 -794.
  • 2Shan Q, Jia J, Agarwala A. High-quality motion deblur- ring from a single image [J]. SIGGRAPH, ACM, 2008: 1 -10.
  • 3Krishnan D, Fergus R. Fastimage deconvolution using hy- per-laplacian [ C .] //Proceedings of Neural Information Processing Systems Blurred Lut Nr, 2009.
  • 4Kotera J, Sroubek F, Milanfar P. Blind deconvolution using alternating maximum a posteriori estimation with heavy-tailed priors [ M ] . Computer Analysis of Images and Patterns Springer Berlin Heidelberg, 2013:59 -66.
  • 5Krishnan D, Tay T, Fergus R. Blind decon~olution usinga normalized sparsity measure [ C ]//IEEE CVPR, 2011 : 233 -240.
  • 6Xu L, Zheng S, Jia J. Unnatural L0 sparse representation for natural image dehlurring [ C]//IEEE CVPR, 2013: 1107 - 1114.
  • 7Goldstein T, Stanley O. The split Bregman method for L1- regularized problems I J]. SIAM Journal on Imaging Sci- ences 2.2, 2009 : 323 - 343.
  • 8Levin A, Yair W. Understanding blind deconvolution algo- rithms [ J]. Pattern Analysis and Machine Intelligence, IEEE Transactions on 33, 2011, (12) : 2354 - 2367.
  • 9Amizic B Rafael M. Katsaggelos. Sparse Bayesian blind image deconvolution with parameter estimation [ J]. EUR- ASIP Journal on Image and Video Processing, 2012, (1) : 1 -15.
  • 10Xu L, Jia J. Two-phase kernel estimation for robust motion deblurring [ C ]// Proceedings of the 11 th European confer- ence on Computer vision: Part I, Heraklion, Crete, Greece, 2010, (9): 157-170.

共引文献4

相关作者

内容加载中请稍等...

相关机构

内容加载中请稍等...

相关主题

内容加载中请稍等...

浏览历史

内容加载中请稍等...
;
使用帮助 返回顶部