Skip to content
Snippets Groups Projects
Commit fb9336b0 authored by even's avatar even
Browse files

Poster: retouch

parent 152a1e6b
No related branches found
No related tags found
No related merge requests found
\begin{tabular}{|l||rcl|rcl|}
\hline
With ADS \& CAT: & \multicolumn{3}{c|}{no} & \multicolumn{3}{c|}{yes} \\
\hline
Detected blurred segments
& 17.06 & $\pm$ & 3.22 & 16.83 & $\pm$ & 3.11 \\
%Detected long (> 40 pixels) blurred segments per image
%& 11.24 & $\pm$ & 1.94 & 11.36 & $\pm$ & 1.97 \\
Undetected input segments
& 0.152 & $\pm$ & 0.43 & \textbf{0.003} & $\pm$ & \textbf{0.05} \\
Precision (\%) : $P = \#(D\cap S)/\#D$
& 80.46 & $\pm$ & 7.22 & \textbf{83.87} & $\pm$ & \textbf{6.04} \\
Recall (\%): $R = \#(D\cap S)/\#S$
& 90.23 & $\pm$ & 3.30 & \textbf{91.15} & $\pm$ & \textbf{2.52} \\
F-measure (\%): $F = 2\times P\times R/(P+R)$
& 84.87 & $\pm$ & 4.42 & \textbf{87.23} & $\pm$ & \textbf{3.59} \\
Thickness difference (pixels)
& 0.70 & $\pm$ & 0.24 & \textbf{0.59} & $\pm$ & \textbf{0.19} \\
Angle difference (degrees)
& 0.61 & $\pm$ & 0.66 & \textbf{0.57} & $\pm$ & \textbf{0.62} \\
\hline
\end{tabular}
......@@ -6,7 +6,7 @@
\usepackage{amssymb}
\usepackage{pst-all}
\usepackage{epsfig}
\usepackage{wallpaper}
%\usepackage{wallpaper}
\usepackage{url}
\usepackage[latin1]{inputenc}
%\usepackage[utf8]{inputenc}
......@@ -25,6 +25,7 @@
\definecolor{RougeSombre}{rgb}{0.8, 0.0, 0.0}
\definecolor{Vert}{rgb}{0.0, 1.0, 0.0}
\definecolor{VertSombre}{rgb}{0.0, 0.5, 0.0}
\definecolor{VertClair}{rgb}{0.9, 1.0, 0.7}
\definecolor{Bleu}{rgb}{0.0, 0.0, 1.0}
\definecolor{BleuSombre}{rgb}{0.0, 0.0, 0.8}
\definecolor{PartColor}{rgb}{0.0, 0.5, 0.0}
......@@ -99,7 +100,7 @@
}
\setlength{\wpYoffset}{-2.5cm}
%\setlength{\wpYoffset}{-2.5cm}
\begin{document}
......@@ -112,7 +113,8 @@
%\wpYoffset{0.2cm}
\CenterWallPaper{1.08}{./Illust/porte2.eps}
%\CenterWallPaper{1.08}{./Illust/porte2.eps}
\pagecolor{VertClair}
\sffamily
\newrgbcolor{lightblue}{0. 0.2 0.20}
......@@ -185,9 +187,6 @@ Thick Line Segment Detection with Fast Directional Tracking \\
\begin{itemize}
\item An efficient tool to detect thick straight lines in
gray-level images with few intuitive parameters. \\
\item Estimated thickness that reflects output line quality
(sharpness, scattering, ...). \\
\item Complete unsupervised detection of all lines in the image. \\
\item Based on a former supervised detector \cite{KerautretEven09} +
two new concepts :
\begin{itemize}
......@@ -195,10 +194,14 @@ Thick Line Segment Detection with Fast Directional Tracking \\
as necessary tracking in a once;
\item {\bf Control of assigned thickness (CAT)}: captures
line thickness as soon as possible, then prevents
outliers insertion.
outliers insertion. \\
\end{itemize}
\item Estimated thickness that reflects output line quality
(sharpness, scattering, ...). \\
\item Complete unsupervised detection of all lines in the image.
\end{itemize}
}
\begin{picture}(5,6)(0,0) \end{picture}
}
\end{minipage}
} &
......@@ -216,23 +219,23 @@ Thick Line Segment Detection with Fast Directional Tracking \\
\section*{Underlying discrete geometry notions}
\begin{minipage}{0.6\textwidth}
\begin{itemize}
\item \textbf{Digital straight line (DSL)} $\mathcal{L}(a,b,c,\nu)$,
\item {\bf Digital straight line (DSL)} $\mathcal{L}(a,b,c,\nu)$,
with $(a,b,c,\nu) \in \mathbb{Z}^4$:
set of points $P(x,y)$ of $\mathbb{Z}^2$ that satisfy :
$0 \leq ax + by - c < \nu$.
\item \textbf{DSL thickness}:
\item {\bf DSL thickness}:
$\mu = \frac{\nu}{max(|a|,|b|)}$ \\
\item \textbf{Blurred segment} $\mathcal{B}$ of assigned thickness
\item {\bf Blurred segment} $\mathcal{B}$ of assigned thickness
$\varepsilon$ \cite{DebledAl06}:
set of points in $\mathbb{Z}^2$ that all belong to a covering
digital straight line $\mathcal{L}$ of thickness
$\mu = \varepsilon$. \\
\item \textbf{Directional scan} $DS$ \cite{KerautretEven09}:
\item {\bf Directional scan (DS)} \cite{KerautretEven09}:
ordered partition restricted to the image domain $\mathcal{I}$
of a thick digital straight line $\mathcal{D}$, called the
\textbf{scan strip}, into scans $S_i$, each of them being a
{\bf scan strip}, into scans $S_i$, each of them being a
segment of a naive line $\mathcal{N}_i$, called a
\textbf{scan line}, orthogonal to $\mathcal{D}$.
{\bf scan line}, orthogonal to $\mathcal{D}$.
\end{itemize}
\end{minipage}
\begin{minipage}{0.05\textwidth}
......@@ -264,15 +267,15 @@ Thick Line Segment Detection with Fast Directional Tracking \\
\section*{Main process workflow}
\input{Illust/workflow3}
\begin{itemize}
\item {\bf Parameters}: initial assigned thickness $\varepsilon_0$, \\
\hspace{2.0cm} final size and sparsity thresholds.
\item {\bf Input}: Manually or automatically set initial stroke $AB$.
\item {\bf Output}: Blurred segment with estimated thickness.
\item {\bf Parameters}: initial assigned thickness $\varepsilon_0$,
final size and sparsity thresholds.
\item {\bf Input}: manually or automatically set initial stroke $AB$.
\item {\bf Output}: blurred segment with estimated thickness.
\item {\bf Initial detection}:
construction and extension of a first blurred segment using
points with highest gradient magnitude found in each scan of
a directional scan built on the input stroke.
\item {\bf Validation tests}:
\item {\bf Validation}:
based on size, sparsity or close orientation to input stroke.
\item {\bf Fine tracking}:
construction and extension of the final blurred segment from
......@@ -280,16 +283,19 @@ Thick Line Segment Detection with Fast Directional Tracking \\
with controlled gradient direction found in each scan of an
adaptive directional scan aligned on the initial blurred segment
direction.
\item {\bf Final tests}: minimal final size and sparsity.
\item {\bf Acceptation}: minimal final size and sparsity.
\end{itemize}
\end{minipage}
} &
\begin{minipage}{0.01\textwidth}
\begin{picture}(1,1)(0,0) \end{picture}
% \begin{pspicture}(1,1)(0,0)
% \psframe[fillstyle=solid,fillcolor=VertSombre](0,0)(100,100)
% \end{pspicture}
\end{minipage} &
\fbox{
\begin{minipage}{0.56\textwidth}
\section*{Adaptive Directional Scan}
\section*{Adaptive Directional Scan (ADS)}
Control of the scan strip based \\
on observed line position $C_{i-1}$, \\
direction $\vec{D}_{i-1}$ and thickness $\mu_{i-1}$.
......@@ -324,7 +330,7 @@ Thick Line Segment Detection with Fast Directional Tracking \\
\put(0,0){\makebox(1,20){Fine tracking with ADS \& CAT}}
\end{picture}
\end{tabular} \\
\section*{Control of Assigned Thickness}
\section*{Control of Assigned Thickness (CAT)}
Adjust $\varepsilon$ to detected BS thickness \\
after $\tau$ steps without $\mu$ growth.
\begin{picture}(1,1)(-180,0)
......@@ -373,20 +379,20 @@ Thick Line Segment Detection with Fast Directional Tracking \\
initial assigned thickness $\varepsilon_0$
\item {\bf Output}: list $\mathcal{L}$ of detected blurred segments
crossing $AB$
\begin{itemize}
\begin{enumerate}
\item Finds and sorts local maxima of gradient magnitude in $AB$
\item For each local maximum:
\begin{itemize}
\item detects a blurred segment in direction $AB^{\perp}$
\begin{enumerate}
\item Detects a blurred segment in direction $AB^{\perp}$
(mask points discarded)
\item If a blurred segment $\mathcal{B}$ is detected:
\begin{itemize}
\begin{enumerate}
\item Updates mask $\mathcal{M}$ with dilated points of
$\mathcal{B}$
\item Adds $\mathcal{B}$ to list $\mathcal{L}$.
\end{itemize}
\end{itemize}
\end{itemize}
\end{enumerate}
\end{enumerate}
\end{enumerate}
\end{itemize}
\end{minipage}
} &
......@@ -417,14 +423,14 @@ Thick Line Segment Detection with Fast Directional Tracking \\
\item {\bf Output}:
list $\mathcal{L}$ of detected blurred segments in
image $\mathcal{I}$
\begin{itemize}
\begin{enumerate}
\item Clears list $\mathcal{L}$ and mask $\mathcal{M}$;
\item Sweeps a vertical stroke $AB$ crossing the whole image,
from left to right border with step $\delta$; \\
At each position, runs a multiple line detection and
adds found blurred segments to list $\mathcal{L}$;
\item Repeat this step with a horizontal stroke.
\end{itemize}
\item Repeat step 2 with a horizontal stroke.
\end{enumerate}
\end{itemize}
\end{minipage}
}
......@@ -493,9 +499,11 @@ Thick Line Segment Detection with Fast Directional Tracking \\
\includegraphics[width=0.3\textwidth]{./Illust/randnew1.eps} \\
Random image & & Without ADS \& CAT & & With ADS \& CAT
\end{tabular}
\input{Illust/statsTable} \\
\begin{center}
\input{Illust/statsTable2}
\end{center}
\begin{picture}(10,10)(0,0) \end{picture} \\
\begin{picture}(10,40)(0,0) \end{picture} \\
{\large \bf Conclusion} \\
\hspace{1.0cm} $\Rightarrow$
Competitive performance compared to other classic detectors. \\
......
0% Loading or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment