The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Publications Copernicus
Articles | Volume XLIV-2/W1-2021
Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLIV-2/W1-2021, 171–176, 2021
Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLIV-2/W1-2021, 171–176, 2021

  15 Apr 2021

15 Apr 2021


V. B. S. Prasath1,2,3,4, N. N. Hien5, D. N. H. Thanh6, and S. Dvoenko7 V. B. S. Prasath et al.
  • 1Division of Biomedical Informatics, Cincinnati Children’s Hospital Medical Center, Cincinnati, OH 45229 USA
  • 2Department of Pediatrics, University of Cincinnati College of Medicine, Cincinnati, OH 45257, USA
  • 3Department of Biomedical Informatics, College of Medicine, University of Cincinnati, OH 45267 USA
  • 4Department of Electrical Engineering and Computer Science, University of Cincinnati, OH 45221, USA
  • 5Dong Thap University, Cao Lanh City, Vietnam
  • 6Department of Information Technology, School of Business Information Technology, University of Economics Ho Chi Minh City, Vietnam
  • 7Institute of Applied Mathematics and Computer Science, Tula State University, Russia

Keywords: Image Restoration, Regularization, Total Variation, Parameter Estimation, Residual Similarity, Convex Minimization

Abstract. Image restoration with regularization models is very popular in the image processing literature. Total variation (TV) is one of the important edge preserving regularization models used, however, to obtain optimal restoration results the regularization parameter needs to be set appropriately. We propose here a new parameter estimation approach for total variation based image restoration. By utilizing known noise levels we compute the regularization parameter by reducing the similarity between residual and noise variances. We use the split Bregman algorithm for the total variation along with this automatic parameter estimation step to obtain a very fast restoration scheme. Experimental results indicate the proposed parameter estimation obtained better denoised images and videos in terms of PSNR and SSIM measures and the computational overload is less compared with other approaches.