• DocumentCode
    3544357
  • Title

    Demosaicing with improved edge direction detection

  • Author

    Wang, Xiaomeng ; Lin, Weisi ; Xue, Ping

  • Author_Institution
    Sch. of Electr. & Electron. Eng., Nanyang Technol. Univ., Singapore
  • fYear
    2005
  • fDate
    23-26 May 2005
  • Firstpage
    2048
  • Abstract
    A new edge adaptive color demosaicing method for Bayer pattern images of single-sensor digital cameras is presented in this paper. An edge direction detector for narrow edges is proposed by making full use of inter-channel correlation to determine edge directions within the smallest detection radius. A combinative criterion is then formulated to cater for the diversity of edge occurrence in real-world scenes. The improvement of the proposed demosaicing scheme is achieved by an effective detection on edge directions, emphasizing on green channel restoration and the overall refinement. Experimental results show that the new scheme preserves better edge details, reduces color aliasing artifacts, and achieves both higher signal fidelity and higher visual image quality as compared with some existing schemes.
  • Keywords
    correlation methods; edge detection; image colour analysis; image reconstruction; interpolation; Bayer pattern images; color aliasing artifact reduction; color image demosaicing; color plane interpolation; edge adaptive demosaicing; edge detail preservation; edge reconstruction; green channel restoration; inter-channel correlation; narrow edge direction detection; real-world scene edge occurrence combinative criterion; single-sensor digital cameras; visual image quality; Detectors; Digital cameras; Filters; Frequency estimation; Humans; Image edge detection; Image reconstruction; Interpolation; Layout; Sensor arrays;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Circuits and Systems, 2005. ISCAS 2005. IEEE International Symposium on
  • Print_ISBN
    0-7803-8834-8
  • Type

    conf

  • DOI
    10.1109/ISCAS.2005.1465020
  • Filename
    1465020