MPEG-4 AVC robust video MPEG-4 AVC robust video watermarking based on QIM and watermarking based on QIM and perceptual masking principles perceptual masking principles Marwen HASNAOUI Marwen HASNAOUI Maher BELHAJ Maher BELHAJ Mihai MITREA Mihai MITREA Françoise PRETEUX Françoise PRETEUX ARTEMIS Department ARTEMIS Department Institut Télécom Institut Télécom Télécom SudParis Télécom SudParis Evry - France Evry - France
22
Embed
MPEG-4 AVC robust video watermarking based on QIM and perceptual masking principles Marwen HASNAOUI Maher BELHAJ Mihai MITREA Françoise PRETEUX ARTEMIS.
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
MPEG-4 AVC robust video MPEG-4 AVC robust video watermarking based on QIM and watermarking based on QIM and perceptual masking principlesperceptual masking principles
Marwen HASNAOUIMarwen HASNAOUI
Maher BELHAJMaher BELHAJ
Mihai MITREAMihai MITREA
Françoise PRETEUXFrançoise PRETEUX
ARTEMIS DepartmentARTEMIS Department Institut Télécom Institut Télécom
Télécom SudParis Télécom SudParis Evry - France Evry - France
Additional constraints for transparent MPEG-4 AVC watermarking in:Additional constraints for transparent MPEG-4 AVC watermarking in:S. Duta, M. MitreaS. Duta, M. Mitrea, F. Preteux, M. Belhaj, “The MPEG-4 AVC domain watermarking , F. Preteux, M. Belhaj, “The MPEG-4 AVC domain watermarking transparency”, transparency”, Proc. SPIE Vol. 6982, April 2008Proc. SPIE Vol. 6982, April 2008
16//116/ xxxx zx
Between myth and realityBetween myth and reality
Unmarked block energy
Marked block energy
The mean of energy of unmarked selected blocks
The standard deviation of energy of
unmarked selected blocks
Mark detection Mark detection
Detection variable
))(())(()( ktzktzQmY AVCt
AVCt
KeyQuantization step
k
Watermarked blockz
Without attack
Between myth and realityBetween myth and reality
Decision
Y(m =d) 2 Decision
Y(m =d+1) 2
dIsup 1inf dI dI inf 1sup dI
01infsup dIdIm
m 1
m
dmI d 2
)2)1(()inf(
m
dmI d 2
)2)1(()sup(
Mark detection Mark detection
After an attack
Detection variable
KeyQuantization step
k
Attacked block
'z
Without attack
))'(())'(()( nAVCt
nAVCt ktzktzQmY
Between myth and realityBetween myth and reality
Decision
Y(m =d) 2 Decision
Y(m =d+1) 2
dIsup 1inf dI dI inf 1sup dI
MEDIEVALS video corpus: MEDIEVALS video corpus: https://www.medialive.com/Medievalshttps://www.medialive.com/Medievals
2 h of video in 2 h of video in HDHD and and SDSD
Different compression rate: 1Mo, 5 Mo, 10 Mo Different compression rate: 1Mo, 5 Mo, 10 Mo
MPEG-4 AVC baseline profile, only level 2, 3 and 4MPEG-4 AVC baseline profile, only level 2, 3 and 4
The watermarking quantization step is chosen according to Golikiri:The watermarking quantization step is chosen according to Golikiri:
New masking model adapted to New masking model adapted to MPEG‑4 AVC featuresMPEG‑4 AVC features
An extension of the first MPEG‑4 AVC robust An extension of the first MPEG‑4 AVC robust watermarking method against transcoding (about 6.5% of watermarking method against transcoding (about 6.5% of errors) and geometric attacks (about 11.57 % of errors)errors) and geometric attacks (about 11.57 % of errors)
Quantization watermarking on Quantization watermarking on compressed domain compressed domain
Information theory modeling attackInformation theory modeling attack
Perspectives
Ameliorate the masking model Ameliorate the masking model
Technical solution: watermarking systemTechnical solution: watermarking system