deletions | additions       diff --git a/Another_idea_if_the_homography__.tex b/Another_idea_if_the_homography__.tex  index 9a94b82..1dd83dc 100644  --- a/Another_idea_if_the_homography__.tex  +++ b/Another_idea_if_the_homography__.tex 
...
Another idea---if the homography-based approach is used for  labeling---was to use the amount of detected keypoints ($K$)  as confidence value. However, again, no linear relation relationship  between $K$ and the error in $x-direction$ ($X$) and  the two variables error in $y-direction$ ($Y$)  could be found (Figure~\ref{fig:cor_sim_confi}).  \begin{figure}[h!]  \begin{center}  \includegraphics[width=0.7\columnwidth]{figures/fig_cor_sift_confi/default_figure}  \caption{{\label{fig:cor_sift_confi} Dependence structure between number of keypoints and quality  of the measurement.%  }}  \end{center}  \end{figure} (Figure~\ref{fig:cor_sim_confi}): $\rho_{K, X} = 0.10$, $\rho_{K, Y} = 0.01$.