• DocumentCode
    152915
  • Title

    Color normalization method using hue histogram Gaussian mixtures

  • Author

    Onder, Devrim

  • fYear
    2014
  • fDate
    23-25 April 2014
  • Firstpage
    1837
  • Lastpage
    1840
  • Abstract
    Color variations occurred due to inconsistencies in histology slides preparation stages degrades the quantitative analysis performance of histopathological images. Several color deconvolution and color modification methods have been proposed to eliminate these negative effects. In this article, a Hue, Saturation, Value space color normalization method for hematoxylin and eosin stained histopathology slide images is presented. In the proposed method, a hue histogram is represented as a binary Gaussian mixture and then each mixture component is transferred to a predefined fixed Gaussian distribution. The proposed method does not require any user intervention. The impact of the proposed method to the separation of the texture feature vector groups obtained from healthy and cancerous tissue sections of a histopathological image database was presented by comparing the group separability measures calculated using original and normalized images.
  • Keywords
    Gaussian distribution; chemical analysis; deconvolution; image colour analysis; image texture; binary Gaussian mixture; color deconvolution methods; color modification methods; color normalization method; color variations; fixed Gaussian distribution; histopathological images; hue histogram Gaussian mixtures; quantitative analysis; texture feature vector groups; Conferences; Histograms; Image color analysis; Microscopy; Nickel; Signal processing; Statistical analysis; color normalization; gaussian mixture; hematoxylin-eosin; histopathology; hue-saturation-value;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Signal Processing and Communications Applications Conference (SIU), 2014 22nd
  • Conference_Location
    Trabzon
  • Type

    conf

  • DOI
    10.1109/SIU.2014.6830610
  • Filename
    6830610