Correlation modeling for compression of computed tomography images

Juan Munoz-Gomez, Joan Bartrina-Rapesta, Michael W. Marcellin, Joan Serra-Sagrista

Research output: Contribution to journalArticleResearchpeer-review

5 Citations (Scopus)

Abstract

Computed tomography (CT) is a noninvasive medical test obtained via a series of X-ray exposures resulting in 3-D images that aid medical diagnosis. Previous approaches for coding such 3-D images propose to employ multicomponent transforms to exploit correlation among CT slices, but these approaches do not always improve coding performance with respect to a simpler slice-by-slice coding approach. In this paper, we propose a novel analysis which accurately predicts when the use of a multicomponent transform is profitable. This analysis models the correlation coefficient r based on image acquisition parameters readily available at acquisition time. Extensive experimental results from multiple image sensors suggest that multicomponent transforms are appropriate for images with correlation coefficient r in excess of 0.87. © 2013 IEEE.
Original languageEnglish
Article number6517882
Pages (from-to)928-935
JournalIEEE Journal of Biomedical and Health Informatics
Volume17
Issue number5
DOIs
Publication statusPublished - 18 Sep 2013

Keywords

  • Computed tomography (CT) image compression
  • JPEG2000 coding standard
  • correlation modeling
  • digital imaging and communications in medicine (DICOM) protocol
  • multicomponent transforms

Fingerprint Dive into the research topics of 'Correlation modeling for compression of computed tomography images'. Together they form a unique fingerprint.

Cite this