• DocumentCode
    242950
  • Title

    Inpainting by an Inverse Problem Resolution

  • Author

    Mezhoud, Djaafer ; Nouri, Fatma Zohra ; Spiteri, Pierre

  • Author_Institution
    Lab. de Modelisation Math. et Simulation Numerique, Univ. Badji Mokhtar, Annaba, Algeria
  • fYear
    2014
  • fDate
    16-18 July 2014
  • Firstpage
    298
  • Lastpage
    301
  • Abstract
    The in painting is a fundamental process in the image processing, which has several applications, such as image restoration, reconstructing damaged parts, the removal of unwanted objects in the image etc. The goal is to produce a revised image where treated parts are perfectly fused in it. In this paper we propose an approach which involves the direct solution of a system of Navier-Stokes equations for an incompressible Newtonian fluid. The main idea is to consider the image intensity as a function of the stream flow. The resolution algorithm automatically transports the information correction to the damaged part of the image, a process which can simultaneously reconstruct several damaged regions by solving an inverse problem, without requiring the user to specify a location treatment. The mathematical analysis of the Navier-Stokes (NS) equations is sufficiently developed, see for example R. Temam (1977), so we try to draw analogies between these equations and in painting which may introduce ideas of fluid mechanics and dynamics in problems of image processing.
  • Keywords
    Navier-Stokes equations; fluid mechanics; image restoration; inverse problems; Navier-Stokes equations; fluid dynamics; fluid mechanics; image intensity; image processing; image reconstruction; image restoration; incompressible Newtonian fluid; information correction; inpainting; inverse problem resolution; revised image; stream flow function; Visualization; Fluid Dynamics; Inpainting; inverse Problems;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Information Visualisation (IV), 2014 18th International Conference on
  • Conference_Location
    Paris
  • ISSN
    1550-6037
  • Type

    conf

  • DOI
    10.1109/IV.2014.12
  • Filename
    6902919