• DocumentCode
    744905
  • Title

    Multiscale contrast enhancement for radiographies: Laplacian pyramid versus fast wavelet transform

  • Author

    Dippel, Sabine ; Stahl, Martin ; Wiemker, Rafael ; Blaffert, Thomas

  • Author_Institution
    Div. Tech. Syst., Philips Res. Lab., Hamburg, Germany
  • Volume
    21
  • Issue
    4
  • fYear
    2002
  • fDate
    4/1/2002 12:00:00 AM
  • Firstpage
    343
  • Lastpage
    353
  • Abstract
    Contrast enhancement of radiographies based on a multiscale decomposition of the images recently has proven to be a far more versatile and efficient method than regular unsharp-masking techniques, while containing these as a subset. In this paper, we compare the performance of two multiscale-methods, namely the Laplacian Pyramid and the fast wavelet transform (FWT). We find that enhancement based on the FWT suffers from one serious drawback-the introduction of visible artifacts when large structures are enhanced strongly. By contrast, the Laplacian Pyramid allows a smooth enhancement of large structures, such that visible artifacts can be avoided. Only for the enhancement of very small details, for denoising applications or compression of images, the FWT may have some advantages over the Laplacian Pyramid.
  • Keywords
    diagnostic radiography; image enhancement; mammography; medical image processing; wavelet transforms; Laplacian Pyramid; denoising applications; fast wavelet transform; images compression; medical diagnostic imaging; multiscale contrast enhancement; very small details; visible artifacts; Biomedical imaging; Computed tomography; Frequency; Image coding; Laplace equations; Mammography; Noise reduction; Radiography; Wavelet transforms; X-ray imaging; Algorithms; Artifacts; Female; Fourier Analysis; Humans; Mammography; Pattern Recognition, Automated; Radiographic Image Enhancement; Sensitivity and Specificity; Signal Processing, Computer-Assisted; Skull;
  • fLanguage
    English
  • Journal_Title
    Medical Imaging, IEEE Transactions on
  • Publisher
    ieee
  • ISSN
    0278-0062
  • Type

    jour

  • DOI
    10.1109/TMI.2002.1000258
  • Filename
    1000258