Changeset 2289 for research/2008displacement
 Timestamp:
 Apr 16, 2008, 12:11:41 AM (12 years ago)
 File:

 1 edited
Legend:
 Unmodified
 Added
 Removed

research/2008displacement/paper/paper.tex
r2277 r2289 102 102 103 103 HVS models are usually lowpass filters. Nasanen \cite{nasanen}, Analoui 104 and Allebach \cite{allebach} found that using gaussian models gave visually104 and Allebach \cite{allebach} found that using Gaussian models gave visually 105 105 pleasing results, an observation confirmed by independent visual perception 106 106 studies \cite{mcnamara}. … … 141 141 \end{equation} 142 142 143 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 144 convolution and $v$ is a model for the human visual system. 145 145 146 146 To compensate for the slight translation experienced in the halftone, we … … 151 151 \end{equation} 152 152 153 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. 155 156 A simple example can be given using a Gaussian HVS model: 157 157 158 158 \begin{equation} … … 166 166 \end{equation} 167 167 168 Experiment showsthat for a given image and a given corresponding halftone,168 Experiments show that for a given image and a given corresponding halftone, 169 169 $E_{dx,dy}$ has a local minimum almost always away from $(dx,dy) = (0,0)$ (Fig. 170 170 \ref{fig:lenamin}). Let $E$ be an error metric where this remains true. We … … 179 179 \input{lenamin} 180 180 \caption{Mean square error for the \textit{Lena} image. $v$ is a simple 181 $11\times11$ gaussian convolution kernel with $\sigma = 1.2$ and181 $11\times11$ Gaussian convolution kernel with $\sigma = 1.2$ and 182 182 $(dx,dy)$ vary in $[1,1]\times[1,1]$.} 183 183 \label{fig:lenamin} … … 191 191 smaller, with the exact same input and output images. 192 192 193 Experiment showsthat the corrected error is always noticeably smaller except193 Experiments show that the corrected error is always noticeably smaller except 194 194 in the case of images that are already mostly pure black and white. The 195 195 experiment was performed on a database of 10,000 images from common computer … … 300 300 301 301 First 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 studying302 error metric $E$ based on a standard Gaussian HVS model. Since we are studying 303 303 algorithms on different images but error values are only meaningful for a given 304 304 image, we chose a Condorcet voting scheme to determine winners. $E_{min}$ is
Note: See TracChangeset
for help on using the changeset viewer.