Changeset 2289 for research


Ignore:
Timestamp:
Apr 16, 2008, 12:11:41 AM (12 years ago)
Author:
Sam Hocevar
Message:
  • Applied changes suggested by reviewer #1: -Page 3: The Latex "\noindent" could be added after equations (1) and (2). -Page 3, paragraph 3: gaussian -> Gaussian -Page 3 (two times): Experiment shows -> Experiments show ??
File:
1 edited

Legend:

Unmodified
Added
Removed
  • research/2008-displacement/paper/paper.tex

    r2277 r2289  
    102102
    103103HVS models are usually low-pass filters. Nasanen \cite{nasanen}, Analoui
    104 and Allebach \cite{allebach} found that using gaussian models gave visually
     104and Allebach \cite{allebach} found that using Gaussian models gave visually
    105105pleasing results, an observation confirmed by independent visual perception
    106106studies \cite{mcnamara}.
     
    141141\end{equation}
    142142
    143 where $w$ and $h$ are the image dimensions, $*$ denotes the convolution and $v$
    144 is a model for the human visual system.
     143\noindent where $w$ and $h$ are the image dimensions, $*$ denotes the
     144convolution and $v$ is a model for the human visual system.
    145145
    146146To compensate for the slight translation experienced in the halftone, we
     
    151151\end{equation}
    152152
    153 where $t_{dx,dy}$ is an operator which translates the image along the $(dx,dy)$
    154 vector.
    155 
    156 A simple example can be given using a gaussian HVS model:
     153\noindent where $t_{dx,dy}$ is an operator which translates the image along the
     154$(dx,dy)$ vector.
     155
     156A simple example can be given using a Gaussian HVS model:
    157157
    158158\begin{equation}
     
    166166\end{equation}
    167167
    168 Experiment shows that for a given image and a given corresponding halftone,
     168Experiments show that for a given image and a given corresponding halftone,
    169169$E_{dx,dy}$ has a local minimum almost always away from $(dx,dy) = (0,0)$ (Fig.
    170170\ref{fig:lena-min}). Let $E$ be an error metric where this remains true. We
     
    179179   \input{lena-min}
    180180   \caption{Mean square error for the \textit{Lena} image. $v$ is a simple
    181             $11\times11$ gaussian convolution kernel with $\sigma = 1.2$ and
     181            $11\times11$ Gaussian convolution kernel with $\sigma = 1.2$ and
    182182            $(dx,dy)$ vary in $[-1,1]\times[-1,1]$.}
    183183   \label{fig:lena-min}
     
    191191smaller, with the exact same input and output images.
    192192
    193 Experiment shows that the corrected error is always noticeably smaller except
     193Experiments show that the corrected error is always noticeably smaller except
    194194in the case of images that are already mostly pure black and white. The
    195195experiment was performed on a database of 10,000 images from common computer
     
    300300
    301301First we studied all possible coefficients on a pool of 250 images with an
    302 error metric $E$ based on a standard gaussian HVS model. Since we are studying
     302error metric $E$ based on a standard Gaussian HVS model. Since we are studying
    303303algorithms on different images but error values are only meaningful for a given
    304304image, we chose a Condorcet voting scheme to determine winners. $E_{min}$ is
Note: See TracChangeset for help on using the changeset viewer.