FASTRELIABLEMULTI-SCALEMOTION REGIONDETECTIONINVIDEOPROCESSING
Lists
Only you see this
Abstract
Motion region detection isanimportant vision topic usually tackled byabackground subtraction principle, which hassomepractical restrictions. Wehence propose amulti-scale motion region detection technique that canfast andreliably segment foreground motion regions fromtwosuccessive video frames. Thekeyidea istoleverage multi-scale structural aggregation toeffectively accentuate real motion changes while suppressing trivial noisy changes. Consequently, this technique canbeeffectively applied tomotion region-of-interest (ROI) based video coding. Ourexperiments showthat theproposed algorithm canreliably extract motion regions andisless sensitive to thresholds thansingle-scale methods. Compared withaH.264/AVC encoder, theproposed semantic video encoder achieves abitrate savingratio ofupto34%atthesimilar video quality, besides anoverall speedup factor of2.6to3.6. Themotion-ROI detection canprocess a352x288size video at20fpsonanIntel Pentium 4processor. IndexTerms-Motion detection, image change detection, video coding, video signal processing, poserecognition
BibTeX
@inproceedings{lu2007fastreliablemulti,
author = {Jiangbo Lu},
title = {{FASTRELIABLEMULTI-SCALEMOTION REGIONDETECTIONINVIDEOPROCESSING}},
booktitle = {IEEE International Conference on Acoustics, Speech and Signal Processing},
year = {2007},
url = {https://openreview.net/forum?id=qNZhgXKudDKC},
}Loading related content...