RMIT University
Browse

Robust Filtering Technique for Reduction of Temporal Fluctuation in H.264 Video Sequences

journal contribution
posted on 2024-11-01, 07:36 authored by Jie Yang, Hong Ren WuHong Ren Wu
The H.264/advanced video coding standard applies a spatial loop-filter in both encoder and decoder to reduce the blocking artifact. However, the temporal fluctuation artifact is still noticeable between intra-coded frames or between an intra (I)-frame and its preceding or subsequent inter-prediction (P) frames. This letter proposes a temporal postfilter using a robust statistical model to reduce the temporal fluctuation and to preserve the motion sharpness of the reconstructed video. The novelty is the implementation of a simplified robust statistical model as well as the Huber statistical model for the temporal fluctuation reduction. Performance evaluations of fluctuation reduction have been conducted using the proposed robust statistical temporal filter (RSTF), the Huber filter, and three other temporal filtering techniques, in terms of a temporal impairment metric (i.e., the sum of squared differences), and a spatial quality metric (i.e., the peak-signal-to-noise ratio). Consistent improvements of the temporal fluctuation reduction and the perceptual picture quality produced by the RSTF are demonstrated using the aforementioned performance criteria, and can be best appreciated by viewing the sequences on professional broadcasting television monitors.

History

Related Materials

  1. 1.
    DOI - Is published in 10.1109/TCSVT.2009.2035850
  2. 2.
    ISSN - Is published in 10518215

Journal

IEEE Transactions On Circuits And Systems For Video Technology

Volume

20

Issue

3

Start page

458

End page

462

Total pages

5

Publisher

IEEE

Place published

United States

Language

English

Copyright

Copyright 2010 IEEE

Former Identifier

2006019435

Esploro creation date

2020-06-22

Fedora creation date

2010-11-19

Usage metrics

    Scholarly Works

    Exports

    RefWorks
    BibTeX
    Ref. manager
    Endnote
    DataCite
    NLM
    DC