JPEG2000 ROI coding method with perfect fine-grain accuracy and lossless recovery

Joan Bartrina-Rapesta*, Joan Serra-Sagristà, Francesc Aulí-Llinàs, Juan Muñoz Gómez

*Corresponding author for this work

Research output: Contribution to journalArticleResearchpeer-review

4 Citations (Scopus)

Abstract

Medical images have high spatial and high bitdepth resolution (12 bits per sample or more). These high resolutions allow computer-aided diagnosis, which are exploited by radiologists to identify relevant medical areas, known as Regions of Interest (ROI). In image compression, ROI coding allows to recover the ROI earlier than the rest of the image. In JPEG2000, ROI coding may be provided through two different mechanisms: either by modifying wavelet coefficients, or by rate-distortion optimization techniques. The former obtains an excellent accuracy to delimit ROIs, but, in some cases, the ROI and the background can not be encoded losslessly; the latter is usually not able to achieve the intended fine-grain accuracy, but it overcomes the lossless encoding shortcoming. This article introduces a ROI coding method based on rate-distortion optimization techniques that recovers the ROI and the background losslessly, regardless of the high bit-depth resolution, and that yields an accuracy equivalent to MaxShift and Scaling, the two compliant JPEG2000 ROI coding methods based on modifying wavelet coefficients. The proposed method is JPEG2000 compliant.

Original languageAmerican English
Pages (from-to)558-562
Number of pages5
JournalConference Record - Asilomar Conference on Signals, Systems and Computers
DOIs
Publication statusPublished - 2009

Keywords

  • Digital mammogram compression
  • JPEG2000 standard
  • Rate-distortion
  • Region of interest coding

Fingerprint Dive into the research topics of 'JPEG2000 ROI coding method with perfect fine-grain accuracy and lossless recovery'. Together they form a unique fingerprint.

Cite this