Automated Removal of quasi-PEriodic NOise using frequency domain Statistics


    The following codes can be used for academic research only:

    MATLAB function: ARPENOS.m v1.2

    Example script and images: ARPENOS_example.m; mandril_gray.tif; boat.png
    Needs symmetrically padded convolution: sym_conv2.m

    Description of the Total Variation minimization postprocessing: note_TVmin.pdf

    Additional experiments (canvas pattern removal in paintings): canvas.m. Images: Rockwell - Carry On - 002.jpg (from here), Rockwell - Carry On - 003.jpg (from here), Rockwell - The Adventure Trail - 003.jpg (from here), Rockwell - Our Heritage - 002.jpg (from here), and one_inclusions.sm_.jpg (from here).

    Please cite:
       F. Sur, M. Grédiac.  Automated removal of quasiperiodic noise using frequency domain statistics. Journal of Electronic Imaging, IS&T / SPIE, vol. 24, no. 1, pages 013003/1-19, IS&T / SPIE, 2015. Link. Pdf file. Bibtex entry.

    Additional experiments can be found in:
       F. Sur, M. Grédiac. An automated approach to quasi-periodic noise removal in natural images. INRIA report RR-8660, January 2015. 57 pages. HAL link. Pdf file. Bibtex entry.

    See also AC-ARPENOS and the corresponding image database.

    January 2015: v1.0, first release.  
    June 2015: v1.1, spectrum inerpolation with TV minimization.
    January 2016: experiments on canvas pattern removal.
    August 2016: v1.2, numerical scheme for constrained TV minimization corrected.

Last modified: August 26th, 2016.