Motion-compensated wavelet video coding using adaptive mode selection

Fan Zhai*, Thrasyvoulos N. Pappas

*Corresponding author for this work

Research output: Contribution to journalConference articlepeer-review


A motion-compensated wavelet video coder is presented that uses adaptive mode selection (AMS) for each macroblock (MB). The block-based motion estimation is performed in the spatial domain, and an embedded zerotree wavelet coder (EZW) is employed to encode the residue frame. In contrast to other motion-compensated wavelet video coders, where all the MBs are forced to be in INTER mode, we construct the residue frame by combining the prediction residual of the INTER MBs with the coding residual of the INTRA and INTER_ENCODE MBs. Different from INTER MBs that are not coded, the INTRA and INTER_ENCODE MBs are encoded separately by a DCT coder. By adaptively selecting the quantizers of the INTRA and INTER_ENCODE coded MBs, our goal is to equalize the characteristics of the residue frame in order to improve the overall coding efficiency of the wavelet coder. The mode selection is based on the variance of the MB, the variance of the prediction error, and the variance of the neighboring MBs' residual. Simulations show that the proposed motion-compensated wavelet video coder achieves a gain of around 0.7-0.8dB PSNR over MPEG-2 TM5, and a comparable PSNR to other 2D motion-compensated wavelet-based video codecs. It also provides potential visual quality improvement.

Original languageEnglish (US)
Pages (from-to)1362-1370
Number of pages9
JournalProceedings of SPIE - The International Society for Optical Engineering
Issue numberPART 2
StatePublished - 2004
EventVisual Communications and Image Processing 2004 - San Jose, CA, United States
Duration: Jan 20 2004Jan 22 2004

ASJC Scopus subject areas

  • Electronic, Optical and Magnetic Materials
  • Condensed Matter Physics
  • Computer Science Applications
  • Applied Mathematics
  • Electrical and Electronic Engineering


Dive into the research topics of 'Motion-compensated wavelet video coding using adaptive mode selection'. Together they form a unique fingerprint.

Cite this