quelques modifs

[canny.git] / experiments.tex

diff --git a/experiments.tex b/experiments.tex
index 498e47285d12a7f5ae95b3f3c2304f0d86375d56..aa49e840c172619692b4e864b8cfaf195acf08d8 100644 (file)
--- a/experiments.tex
+++ b/experiments.tex
@@ -42,7 +42,7 @@ $$
 \hline
 \end{array}
 $$
 \hline
 \end{array}
 $$

-\caption{Matrix Generator for $\hat{H}$ in STC}\label{table:matrices:H}
+\caption{Matrix Generator for $\hat{H}$ in STC.}\label{table:matrices:H}

 \end{table}
 
 
 \end{table}
 
 


@@ -62,80 +62,80 @@ 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}, 
-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 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.
+% 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}
-\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$6.35\%\\ 
-\hline
-PSNR & 61.86 & 63.48 &  66.55  &  63.7  & 64.65 & {67.08} & 60.8 & 62.9&65.9 & 68.3 & 65.8 & 69.2\\ 
-\hline
-PSNR-HVS-M & 72.9 & 75.39 & 78.6  & 75.5  & 76.67 & {79.6} & 71.8  & 76.0 &
-76.7 & 80.35 & 77.6 & 81.2 \\
-\hline
-wPSNR & 77.47 & 80.59 & 86.43& 86.28  & 83.03 & {88.6} & 76.7 & 83& 83.8 & 90.4 & 85.2 & 91.9\\ 
-\hline
-\end{tabular}
-\end{small}
-\end{center}
-\caption{Quality measures of steganography approaches\label{table:quality}}
-\end{table*}
+% \begin{table*}
+% \begin{center}
+% \begin{small}
+% \setlength{\tabcolsep}{3pt}
+% \begin{tabular}{|c|c|c||c|c|c|c|c|c|c|c|c|c|}
+% \hline
+% Schemes & \multicolumn{4}{|c|}{STABYLO} & \multicolumn{2}{|c|}{HUGO}& \multicolumn{2}{|c|}{EAISLSBMR} & \multicolumn{2}{|c|}{WOW} & \multicolumn{2}{|c|}{UNIWARD}\\
+% \hline
+% Embedding &   Fixed & \multicolumn{3}{|c|}{Adaptive (about 6.35\%)} &  Fixed &Adaptive & Fixed &Adaptive & Fixed &Adaptive & Fixed &Adaptive \\
+% \hline
+% Rate &   10\% &  + sample &  +STC(7) & +STC(6) &  10\%&$\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*}

 
 
 
 
 
 

-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}.
+% 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. 

 
 
 
 

-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.  
+% 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.  

 
 
 
 

-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.
+% 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. 

 
 
 
 
 
 


@@ -150,9 +150,12 @@ The steganalysis quality of our approach has been evaluated through the % two
 Ensemble Classifier~\cite{DBLP:journals/tifs/KodovskyFH12} based steganalyser.
 Its  particularization to spatial domain is 
 considered as state of the art steganalysers.
 Ensemble Classifier~\cite{DBLP:journals/tifs/KodovskyFH12} based steganalyser.
 Its  particularization to spatial domain is 
 considered as state of the art steganalysers.

-Firstly, a space 
-of 686 co-occurrence and Markov features is extracted from the 
-set of cover images and the set of training images. Next a small 
+Features that are embedded into this steganalysis process 
+are CCPEV and SPAM features as described 
+in~\cite{DBLP:dblp_conf/mediaforensics/KodovskyPF10}.
+They  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 
 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 


@@ -190,7 +193,7 @@ Ensemble Classifier & 0.35 & 0.44 & 0.47 & 0.47     & 0.48 &  0.49  &  0.43  & 0
 \end{tabular}
 \end{small}
 \end{center}
 \end{tabular}
 \end{small}
 \end{center}

-\caption{Steganalysing STABYLO\label{table:steganalyse}} 
+\caption{Steganalysing STABYLO\label{table:steganalyse}.} 

 \end{table*}
 
 
 \end{table*}
 
 


@@ -205,11 +208,29 @@ as far as we know.
 However by combining Adaptive and STC strategies
 our approach obtains similar results than the ones of these schemes.
 
 However by combining Adaptive and STC strategies
 our approach obtains similar results than the ones of these schemes.
 

-Compared to EAILSBMR, we obtain better results when the strategy is 
+Compared to EAILSBMR, we obtain similar
+results when the strategy is 

 Adaptive. 
 Adaptive. 

-However due to its 
-huge number of integration features, it is not lightweight.
+However due to its huge number of integration features, it is not lightweight.

 
 All these numerical experiments confirm 
 the objective presented in the motivations:
 
 All these numerical experiments confirm 
 the objective presented in the motivations:

-providing an efficient steganography approach in a lightweight manner.
+providing an efficient steganography approach in a lightweight manner
+for small payload.
+
+In Figure~\ref{fig:error}, 
+Ensemble Classifier has been used with all the previous 
+steganographic schemes with 4 different payloads.
+It can be observed that face to high values of payload, 
+STABYLO is definitely not secure enough.
+However thanks to an efficient very low-complexity (Fig.\ref{fig:compared}), 
+we argue that the user should embed tiny messages in many images 
+than a larger message in only one image.
+\begin{figure}
+\begin{center}
+\includegraphics[scale=0.5]{error}
+\end{center}
+\caption{Testing errors obtained by Ensemble classifier with 
+WOW/UNIWARD, HUGO, and STABYLO w.r.t. payload.}
+\label{fig:error} 
+\end{figure}