Compressing moving pictures using the APEX neural principal component extractor

K. I. Diamantaras, S. Y. Kung

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

In this paper we propose an application of the optimal Karhunen-Lokve Transform in place of the traditional DCT for compressing Intra-frames in the MPEG protocol. The I-frames attain the smallest compression ratio since they are coded without reference to any other frames. The difficulty of KLT that additional bit-rate is required to make the image-dependent transform basis known to the decoder is overcome by using the KLT basis of the previous I-or P-frame which we argue, is very similar to the basis of the current frame. The previous frame is already known to the decoder, therefore no additional information needs to be sent out. Paying attention to the non-stationary statistics of images we also propose to split the I-frames in N parts and use a dedicated transform basis for each part. Since the KLT basis must be updated continuously we make use of the adaptive Principal Component extractor network (APEX) that can incrementally estimate the new basis for the next frame. Our simulations for two sports sequences indicate that a 10% to 30% error improvement is achieved using the same number of components for DCT and KLT. Furthermore, the MPEG bit-stream format does not have to change to accommodate the proposed technique.

Original languageEnglish (US)
Title of host publicationNeural Networks for Signal Processing III - Proceedings of the 1993 IEEE Workshop, NNSP 1993
EditorsC.A. Kamm, G.M. Kuhn, R. Chellappa, B. Yoon, S.Y. Kung
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages321-330
Number of pages10
ISBN (Electronic)0780309286, 9780780309289
DOIs
StatePublished - 1993
Event1993 3rd IEEE-SP Workshop on Neural Networks for Signal Processing, NNSP 1993 - Linthicum Heights, United States
Duration: Sep 6 1993Sep 9 1993

Publication series

NameNeural Networks for Signal Processing III - Proceedings of the 1993 IEEE Workshop, NNSP 1993

Conference

Conference1993 3rd IEEE-SP Workshop on Neural Networks for Signal Processing, NNSP 1993
Country/TerritoryUnited States
CityLinthicum Heights
Period9/6/939/9/93

All Science Journal Classification (ASJC) codes

  • Artificial Intelligence
  • Signal Processing
  • Safety, Risk, Reliability and Quality

Fingerprint

Dive into the research topics of 'Compressing moving pictures using the APEX neural principal component extractor'. Together they form a unique fingerprint.

Cite this