Error propagation and distortion modeling in loss-affected predictive video coding

Research output: Chapter in Book/Report/Conference proceedingChapter

1 Scopus citations


The highly complex prediction dependency structure employed in current video coding algorithms makes the resulting compressed bitstreams highly susceptible to data loss or corruption. Caused by transmission over imperfect communication channels or faulty storage devices these errors propagate then into other segments of the video bitstream thereby causing wild variations in quality of the reconstructed content. This chapter reviews the state-of-the-art in modeling the above error propagation phenomenon in predictive video coding and the resulting increase in video distortion across the affected media presentation. We focus in greater detail on the most important recent advances in packet-based distortion estimation techniques and examine some of the most interesting related discoveries.We show that video distortion is not only affected by the amount of data lost but also by the spatio-temporal distribution of the affected data segments. Furthermore, we illustrate cases where contrary to common belief subsequent packet loss actually leads to a reduction in video distortion and where surprisingly increased burstiness of the loss process again contributes to a smaller drop in video quality.

Original languageEnglish (US)
Title of host publicationMultimedia Analysis, Processing and Communications
EditorsWeisi Lin, Dacheng Tao, Janusz Kacprzyk, Zhu Li, Ebroul Izquierdo, Haohong Wang
Number of pages40
StatePublished - 2011
Externally publishedYes

Publication series

NameStudies in Computational Intelligence
ISSN (Print)1860-949X

All Science Journal Classification (ASJC) codes

  • Artificial Intelligence


  • Markov models
  • burst packet loss errors
  • distortion modeling
  • error propagation
  • lossy transmission
  • packet loss
  • video coding


Dive into the research topics of 'Error propagation and distortion modeling in loss-affected predictive video coding'. Together they form a unique fingerprint.

Cite this