Merge branch 'master' of ssh://info.iut-bm.univ-fcomte.fr/canny

[canny.git] / experiments.tex

diff --git a/experiments.tex b/experiments.tex
index 6660072af5410e4ad0c771183a3c0eb63cd8c17d..9590b1c4982484f286f04cd151f3b821449ac6c5 100644 (file)
--- a/experiments.tex
+++ b/experiments.tex
@@ -1,3 +1,5 @@
+First of all, the whole code of STABYLO can be downloaded
+\footnote{\url{http://http://members.femto-st.fr/jf-couchot/en/stabylo}}.

 For all the experiments, the whole set of 10,000 images 
 of the BOSS contest~\cite{Boss10} database is taken. 
 In this set, each cover is a $512\times 512$
 For all the experiments, the whole set of 10,000 images 
 of the BOSS contest~\cite{Boss10} database is taken. 
 In this set, each cover is a $512\times 512$


@@ -10,7 +12,7 @@ We use the matrices $\hat{H}$
 generated by the integers given
 in Table~\ref{table:matrices:H}
 as introduced in~\cite{FillerJF11}, since these ones have experimentally 
 generated by the integers given
 in Table~\ref{table:matrices:H}
 as introduced in~\cite{FillerJF11}, since these ones have experimentally 

-be proven to have the best modification efficiency.
+be proven to have the strongest modification efficiency.

 For instance if the rate between the size of the message and the size of the 
 cover vector
 is 1/4, each number in $\{81, 95, 107, 121\}$ is translated into a binary number 
 For instance if the rate between the size of the message and the size of the 
 cover vector
 is 1/4, each number in $\{81, 95, 107, 121\}$ is translated into a binary number 


@@ -44,112 +46,96 @@ $$
 \end{table}
 
 
 \end{table}
 
 

-Our approach is always compared to Hugo~\cite{DBLP:conf/ih/PevnyFB10}
-and to EAISLSBMR~\cite{Luo:2010:EAI:1824719.1824720}.
-The former is the least detectable information hiding tool in spatial domain 
-and the latter is the work that is the closest to ours, as far as we know. 
-
-
-
-First of all,  in our experiments and with the adaptive scheme, 
-the average size of the message that can be embedded is 16,445 bits.
-It corresponds to an  average payload of 6.35\%. 
-The two other tools will then be compared with this payload. 
-Sections~\ref{sub:quality} and~\ref{sub:steg} respectively present 
-the quality analysis and the security of our scheme. 
+Our approach is always compared to HUGO, to EAISLSBMR, to WOW and to UNIWARD
+for the two strategies Fixed and Adaptive. 
+For the former one, the payload has been set to 10\%. 
+For the latter one, the Canny parameter $T$ has been set to 3. 
+When $b$ is 7, the average size of the message that can be embedded
+is 16,445 bits, 
+that corresponds to an  average payload of 6.35\%.
+For each cover image the STABYLO's embedding rate with these two parameters 
+is memorized. 
+Next each steganographic scheme is executed to produce the stego content of 
+this cover with respect to this embedding rate.  

 
 
 
  
 
 
 
 
  
 

-\subsection{Image quality}\label{sub:quality}
-The visual quality of the STABYLO scheme is evaluated in this section.
-For the sake of completeness, three metrics are computed in these experiments: 
-the Peak Signal to Noise Ratio (PSNR), 
-the PSNR-HVS-M family~\cite{psnrhvsm11}, 
-%the BIQI~\cite{MB10}, 
-and 
-the weighted PSNR (wPSNR)~\cite{DBLP:conf/ih/PereiraVMMP01}.
-The first one is widely used but does not take into
-account the Human Visual System (HVS).
-The other ones have been designed to tackle this problem.
+% \subsection{Image quality}\label{sub:quality}
+% The visual quality of the STABYLO scheme is evaluated in this section.
+% For the sake of completeness, three metrics are computed in these experiments: 
+% the Peak Signal to Noise Ratio (PSNR), 
+% the PSNR-HVS-M family~\cite{psnrhvsm11}, 
+% and 
+% the weighted PSNR (wPSNR)~\cite{DBLP:conf/ih/PereiraVMMP01}.
+% The first one is widely used but does not take into
+% account the Human Visual System (HVS).
+% The other ones have been designed to tackle this problem.

 
 

-If we apply them on the running example, 
-the PSNR, PSNR-HVS-M, and wPSNR values are respectively equal to 
-68.39, 79.85, and 89.71 for the stego Lena when $b$ is equal to 7.
-If $b$ is 6, these values are respectively equal to 
-65.43, 77.2, and 89.35.
+% If we apply them on the running example with the Adaptive and STC strategies, 
+% the PSNR, PSNR-HVS-M, and wPSNR values are respectively equal to 
+% 68.39, 79.85, and 89.71 for the stego Lena when $b$ is equal to 7.
+% If $b$ is 6, these values are respectively equal to 
+% 65.43, 77.2, and 89.35.

 
 
 
 
 
 
 
 

-\begin{table*}
-\begin{center}
-\begin{small}
-\begin{tabular}{|c|c|c||c|c|c|c|c|c|c|c|c|c|}
-\hline
-Schemes & \multicolumn{4}{|c|}{STABYLO} & \multicolumn{2}{|c|}{HUGO}& \multicolumn{2}{|c|}{EAISLSBMR} & \multicolumn{2}{|c|}{WOW} & \multicolumn{2}{|c|}{UNIWARD}\\
-\hline
-Embedding &   Fixed & \multicolumn{3}{|c|}{Adaptive (about 6.35\%)} &  Fixed &Adaptive & Fixed &Adaptive & Fixed &Adaptive & Fixed &Adaptive \\
-\hline
-Rate &   10\% &  + sample &  +STC(7) & +STC(6) &  10\%&6.35\%& 10\%&6.35\%& 10\%&6.35\%& 10\%&6.35\%\\ 
-\hline
-PSNR & 61.86 & 63.48 &  66.55 (\textbf{-0.8\%}) &  63.7  & 64.65 & {67.08} & 60.8 & 62.9&65.9 & 68.3 & 65.8 & 69.2\\ 
-\hline
-PSNR-HVS-M & 72.9 & 75.39 & 78.6 (\textbf{-0.8\%}) & 75.5  & 76.67 & {79.23} & 71.8  & 74.3\\ 
-%\hline
-%BIQI & 28.3 & 28.28 & 28.4 & 28.28 & 28.28 & 28.2 & 28.2\\ 
-\hline
-wPSNR & 77.47 & 80.59 & 86.43(\textbf{-1.6\%})& 86.28  & 83.03 & {88.6} & 76.7 & 83& 83.8 & 90.4 & 85.2 & 91.9\\ 
-\hline
-\end{tabular}
-\end{small}
-\begin{footnotesize}
-\vspace{2em}
-Variances given in bold font express the quality differences between 
-HUGO and STABYLO with  STC+adaptive parameters.
-\end{footnotesize}
+% \begin{table*}
+% \begin{center}
+% \begin{small}
+% \setlength{\tabcolsep}{3pt}
+% \begin{tabular}{|c|c|c||c|c|c|c|c|c|c|c|c|c|}
+% \hline
+% Schemes & \multicolumn{4}{|c|}{STABYLO} & \multicolumn{2}{|c|}{HUGO}& \multicolumn{2}{|c|}{EAISLSBMR} & \multicolumn{2}{|c|}{WOW} & \multicolumn{2}{|c|}{UNIWARD}\\
+% \hline
+% Embedding &   Fixed & \multicolumn{3}{|c|}{Adaptive (about 6.35\%)} &  Fixed &Adaptive & Fixed &Adaptive & Fixed &Adaptive & Fixed &Adaptive \\
+% \hline
+% Rate &   10\% &  + sample &  +STC(7) & +STC(6) &  10\%&$\approx$6.35\%& 10\%&$\approx$6.35\%& 10\%&$\approx$6.35\%& 10\%&$\approx$6.35\%\\ 
+% \hline
+% PSNR & 61.86 & 63.48 &  66.55  &  63.7  & 64.65 & {67.08} & 60.8 & 62.9&65.9 & 68.3 & 65.8 & 69.2\\ 
+% \hline
+% PSNR-HVS-M & 72.9 & 75.39 & 78.6  & 75.5  & 76.67 & {79.6} & 71.8  & 76.0 &
+% 76.7 & 80.35 & 77.6 & 81.2 \\
+% \hline
+% wPSNR & 77.47 & 80.59 & 86.43& 86.28  & 83.03 & {88.6} & 76.7 & 83& 83.8 & 90.4 & 85.2 & 91.9\\ 
+% \hline
+% \end{tabular}
+% \end{small}
+% \end{center}
+% \caption{Quality measures of steganography approaches\label{table:quality}}
+% \end{table*}

 
 

-\end{center}
-\caption{Quality measures of steganography approaches\label{table:quality}}
-\end{table*}

 
 
 
 

+% Results are summarized in Table~\ref{table:quality}.
+% In this table, STC(7) stands for embedding data in the LSB whereas
+% in STC(6), data are hidden in the last two  significant bits. 

 
 

-Results are summarized in Table~\ref{table:quality}.
-Let us give an interpretation of these experiments.
-First of all, the adaptive strategy produces images with lower distortion 
-than the images resulting from the 10\% fixed strategy.
-Numerical results are indeed always greater for the former strategy than 
-for the latter one.
-These results are not surprising since the adaptive strategy aims at 
-embedding messages whose length is decided according to an higher threshold
-into the edge detection.  
-Let us focus on the quality of HUGO images: with a given fixed 
-embedding rate (10\%), 
-HUGO always produces images whose quality is higher than the STABYLO's one.
-However our approach is always better than EAISLSBMR since this one may modify 
-the two least significant bits.

 
 

-If we combine \emph{adaptive} and \emph{STC} strategies 
-(which leads to an average embedding rate equal to 6.35\%)
-our approach  provides metrics equivalent to those provided by HUGO.
-In this column STC(7) stands for embedding data in the LSB whereas
-in STC(6), data are hidden in the last two  significant bits. 
+% Let us give an interpretation of these experiments.
+% First of all, the Adaptive strategy produces images with lower distortion 
+% than the images resulting from the 10\% fixed strategy.
+% Numerical results are indeed always greater for the former strategy than 
+% for the latter one.
+% These results are not surprising since the Adaptive strategy aims at 
+% embedding messages whose length is decided according to a higher threshold
+% into the edge detection.  

 
 
 
 

-
-The quality variance between HUGO and STABYLO for these parameters
-is given in bold font. It is always close to 1\% which confirms 
-the objective presented in the motivations:
-providing an efficient steganography approach in a lightweight manner.
+% If we combine Adaptive and STC strategies 
+% the STABYLO scheme  provides images whose quality is higher than 
+% the EAISLSBMR's one but lower than the quality of high complexity 
+% schemes. Notice that the quality of the less respectful scheme (EAILSBMR) 
+% is lower than 6\% than the one of the most one.

 
 
 
 

-Let us now compare the STABYLO approach with other edge based steganography
-approaches, namely~\cite{DBLP:journals/eswa/ChenCL10,Chang20101286}.
-These two schemes focus on increasing the
-payload while the PSNR is acceptable, but do not 
-give quality metrics for fixed embedding rates from a large base of images. 
+% % Let us now compare the STABYLO approach with other edge based steganography
+% % approaches, namely~\cite{DBLP:journals/eswa/ChenCL10,Chang20101286}.
+% % These two schemes focus on increasing the
+% % payload while the PSNR is acceptable, but do not 
+% % give quality metrics for fixed embedding rates from a large base of images. 

 
 
 
 
 
 


@@ -162,9 +148,20 @@ The steganalysis quality of our approach has been evaluated through the % two
 % AUMP~\cite{Fillatre:2012:ASL:2333143.2333587}
 % and
 Ensemble Classifier~\cite{DBLP:journals/tifs/KodovskyFH12} based steganalyser.
 % AUMP~\cite{Fillatre:2012:ASL:2333143.2333587}
 % and
 Ensemble Classifier~\cite{DBLP:journals/tifs/KodovskyFH12} based steganalyser.

-This approach  aims at detecting hidden bits in grayscale natural
-images and is 
-considered as state of the art steganalysers in the spatial domain~\cite{FK12}.
+Its  particularization to spatial domain is 
+considered as state of the art steganalysers.
+\JFC{Features that are embedded into this steganalysis process 
+are CCPEV and SPAM features as described 
+in~\cite{DBLP:dblp_conf/mediaforensics/KodovskyPF10}.
+These latter are extracted from the 
+set of cover images and the set of training images.}
+Next a small 
+set of weak classifiers is randomly built,
+each one working on a subspace of all the features.
+The final classifier is constructed by a majority voting 
+between the decisions of these  individual classifiers.
+
+

 %The former approach is based on a simplified parametric model of natural images.
 % Parameters are firstly estimated and an adaptive Asymptotically Uniformly Most Powerful
 % (AUMP) test is designed (theoretically and practically), to check whether
 %The former approach is based on a simplified parametric model of natural images.
 % Parameters are firstly estimated and an adaptive Asymptotically Uniformly Most Powerful
 % (AUMP) test is designed (theoretically and practically), to check whether


@@ -179,13 +176,14 @@ considered as state of the art steganalysers in the spatial domain~\cite{FK12}.
 \begin{table*}
 \begin{center}
 \begin{small}
 \begin{table*}
 \begin{center}
 \begin{small}

+\setlength{\tabcolsep}{3pt}

 \begin{tabular}{|c|c|c|c|c|c|c|c|c|c|c|c|c|}
 \hline
 Schemes & \multicolumn{4}{|c|}{STABYLO} & \multicolumn{2}{|c|}{HUGO}& \multicolumn{2}{|c|}{EAISLSBMR} &  \multicolumn{2}{|c|}{WOW} &  \multicolumn{2}{|c|}{UNIWARD}\\
 \hline
 Embedding & Fixed &   \multicolumn{3}{|c|}{Adaptive (about 6.35\%)}  & Fixed & Adapt. & Fixed & Adapt. & Fixed & Adapt. & Fixed & Adapt. \\
 \hline
 \begin{tabular}{|c|c|c|c|c|c|c|c|c|c|c|c|c|}
 \hline
 Schemes & \multicolumn{4}{|c|}{STABYLO} & \multicolumn{2}{|c|}{HUGO}& \multicolumn{2}{|c|}{EAISLSBMR} &  \multicolumn{2}{|c|}{WOW} &  \multicolumn{2}{|c|}{UNIWARD}\\
 \hline
 Embedding & Fixed &   \multicolumn{3}{|c|}{Adaptive (about 6.35\%)}  & Fixed & Adapt. & Fixed & Adapt. & Fixed & Adapt. & Fixed & Adapt. \\
 \hline

-Rate & 10\% &  + sample &   +STC(7) & +STC(6)   & 10\%& 6.35\%& 10\%& 6.35\% & 10\%& 6.35\%& 10\%& 6.35\%\\ 
+Rate & 10\% &  + sample &   +STC(7) & +STC(6)   & 10\%& $\approx$6.35\%& 10\%& $\approx$6.35\% & 10\%& $\approx$6.35\%& 10\%& $\approx$6.35\%\\ 

 \hline
 %AUMP & 0.22 & 0.33 & 0.39  &   0.45    &  0.50 &  0.50 & 0.49 & 0.50 \\
 %\hline
 \hline
 %AUMP & 0.22 & 0.33 & 0.39  &   0.45    &  0.50 &  0.50 & 0.49 & 0.50 \\
 %\hline


@@ -199,20 +197,32 @@ Ensemble Classifier & 0.35 & 0.44 & 0.47 & 0.47     & 0.48 &  0.49  &  0.43  & 0
 \end{table*}
 
 
 \end{table*}
 
 

-Results are summarized in Table~\ref{table:steganalyse}.
+Results of average testing errors 
+are summarized in Table~\ref{table:steganalyse}.

 First of all, STC outperforms the sample strategy %for % the two steganalysers
  as 
 already noticed in the quality analysis presented in the previous section. 
 First of all, STC outperforms the sample strategy %for % the two steganalysers
  as 
 already noticed in the quality analysis presented in the previous section. 

-Next, our approach is more easily detectable than HUGO, which
-is the most secure steganographic tool, as far as we know. 
-However by combining \emph{adaptive} and \emph{STC} strategies
-our approach obtains similar results to HUGO ones.
-
-%%%%et pour b= 6 ?
-
+Next, our approach is more easily detectable than HUGO, 
+WOW and UNIWARD which are the most secure steganographic tool, 
+as far as we know. 
+However by combining Adaptive and STC strategies
+our approach obtains similar results than the ones of these schemes.
+
+Compared to EAILSBMR, we obtain similar
+results when the strategy is 
+Adaptive. 
+However due to its huge number of integration features, it is not lightweight.
+
+All these numerical experiments confirm 
+the objective presented in the motivations:
+providing an efficient steganography approach in a lightweight manner.

 
 

-Compared to EAILSBMR, we obtain better results when the strategy is 
-\emph{adaptive}. 
-However due to its 
-huge number of integration features, it is not lightweight, which justifies 
-in the authors' opinion the consideration of the proposed method.   
+\RC{In Figure~\ref{fig:error}, Ensemble Classifier has been used with all the previsou steganalizers with 3 different payloads. It can be observed that with important payload, STABYLO is not efficient, but as mentionned its complexity is far more simple compared to other tools.\\
+\begin{figure}
+\begin{center}
+\includegraphics[scale=0.5]{error}
+\end{center}
+\caption{Error obtained by Ensemble classifier with WOW/UNIWARD, HUGO, and STABYLO and different paylaods.}
+\label{fig:error} 
+\end{figure}
+}