Multiple description image/video compression using oversampling and noise shaping in the DCT-domain

Michal Palgy, Jan Østergaard, Ram Zamir

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Multiple Descriptions (MDs) source coding is a method to overcome unexpected information loss in a diversity system such as the internet, or a wireless network. In case that the information can suffer loss and still be intelligible, like audio or video, MD is a way to take full advantage of the diversity in the system and provide graceful degradation. Standards-compliant MD coding targeting widely used compression standards, such as JPEG and MPEG, have so far however received little attention in the literature. A recently proposed scheme of MD coding by dithered Delta-Sigma Quantization (DSQ) utilizes oversampling in order to accomplish diversity, and noise shaping to control the distortion-versus-loss trade off at any fixed rate. In this work we implement a DSQ-based MD compression scheme for images/videos working in the discrete cosine transform (DCT) domain. The MD encoder and decoder consist of a standard JPEG/MPEG core, with additional processing and control. We show promising numerical results under various scenarios.

Original languageEnglish
Title of host publication2010 IEEE 26th Convention of Electrical and Electronics Engineers in Israel, IEEEI 2010
Pages965-969
Number of pages5
DOIs
StatePublished - 2010
Event2010 IEEE 26th Convention of Electrical and Electronics Engineers in Israel, IEEEI 2010 - Eilat, Israel
Duration: 17 Nov 201020 Nov 2010

Publication series

Name2010 IEEE 26th Convention of Electrical and Electronics Engineers in Israel, IEEEI 2010

Conference

Conference2010 IEEE 26th Convention of Electrical and Electronics Engineers in Israel, IEEEI 2010
Country/TerritoryIsrael
CityEilat
Period17/11/1020/11/10

Keywords

  • DCT
  • Dithered delta-sigma quantization
  • JPEG
  • MPEG
  • Multiple-description coding

Fingerprint

Dive into the research topics of 'Multiple description image/video compression using oversampling and noise shaping in the DCT-domain'. Together they form a unique fingerprint.

Cite this