<html>
<head>
<meta content="text/html; charset=ISO-8859-1" http-equiv="Content-Type">
<title>mmdcalc</title>
<link href="../tbxdok.css" rel="stylesheet">
</head>
<body>
<table class="topNav">
<tr>
<td class="index">
[<a href="../mmdemos/mmdbrain.html"><tt>mmdbrain</tt></a>]
[<a href="index.html">Up</a>]
[<a href="../mmdemos/mmdcells.html"><tt>mmdcells</tt></a>]
</td>
<td class="title">Demonstrations</td>
</tr>
</table>
<h1>mmdcalc
<br>
<span class="subtitle">Extract the keys of a calculator.
</span>
</h1>
<div class="descr">
<H2>Description</H2>
<div class="H2">
<p>
This procedure detects the keys in a gray-scale image of a calculator. The approach used is watershed Beucher's paradigm. The internal markers are the numbers on the keys and the external markers are obtained by the skiz of the keys background.
</p>
</div>
</div>
<div class="script">
<H2>Demo Script</H2>
<div class="H2">
<div class="slide">
<H3>Reading</H3>
<div class="H3">
<p>
<p>
The gray-scale image of the calculator is read.
</p>
<div class="example">
<div class="listing">
<pre class="user">>>> a = mmreadgray('keyb.tif');</pre>
<pre class="computer"></pre>
<pre class="user">>>> mmshow(a);</pre>
<pre class="computer"></pre>
</div>
<table class="images">
<tbody align="center">
<tr class="image" valign="bottom">
<td><img width="295" src="../images/img_mmdcalc_001.jpg"></td>
<td class="spare"></td>
</tr>
<tr class="title" valign="baseline">
<td><a href="../images/img_mmdcalc_001.jpg">a</a></td>
<td class="spare"></td>
</tr>
</tbody>
</table>
</div>
</p>
</div>
</div>
<div class="slide">
<H3>Gradient image</H3>
<div class="H3">
<p>
<p>
The image edges are enhanced by the gradient operator.
</p>
<div class="example">
<div class="listing">
<pre class="user">>>> b = mmgradm(a, mmsebox());</pre>
<pre class="computer"></pre>
<pre class="user">>>> mmshow(b);</pre>
<pre class="computer"></pre>
</div>
<table class="images">
<tbody align="center">
<tr class="image" valign="bottom">
<td><img width="295" src="../images/img_mmdcalc_002.jpg"></td>
<td class="spare"></td>
</tr>
<tr class="title" valign="baseline">
<td><a href="../images/img_mmdcalc_002.jpg">b</a></td>
<td class="spare"></td>
</tr>
</tbody>
</table>
</div>
</p>
</div>
</div>
<div class="slide">
<H3>Inside marker: Opening top-hat</H3>
<div class="H3">
<p>
<p>
The opening top-hat procedure enhances the small objects relatively to its background. In the calculator image, the digits are enhanced.
</p>
<div class="example">
<div class="listing">
<pre class="user">>>> c = mmopenth(a,mmsebox(5));</pre>
<pre class="computer"></pre>
<pre class="user">>>> mmshow(c);</pre>
<pre class="computer"></pre>
</div>
<table class="images">
<tbody align="center">
<tr class="image" valign="bottom">
<td><img width="295" src="../images/img_mmdcalc_003.jpg"></td>
<td class="spare"></td>
</tr>
<tr class="title" valign="baseline">
<td><a href="../images/img_mmdcalc_003.jpg">c</a></td>
<td class="spare"></td>
</tr>
</tbody>
</table>
</div>
</p>
</div>
</div>
<div class="slide">
<H3>Inside marker: Thresholding </H3>
<div class="H3">
<p>
<p>
The threshold operator is used to separated the enhanced objects. This procedure is quite robust, since the background was reduced to very low levels with the opening top-hat.
</p>
<div class="example">
<div class="listing">
<pre class="user">>>> d = mmthreshad(c, uint8(150));</pre>
<pre class="computer"></pre>
<pre class="user">>>> mmshow(d);</pre>
<pre class="computer"></pre>
</div>
<table class="images">
<tbody align="center">
<tr class="image" valign="bottom">
<td><img width="295" src="../images/img_mmdcalc_004.jpg"></td>
<td class="spare"></td>
</tr>
<tr class="title" valign="baseline">
<td><a href="../images/img_mmdcalc_004.jpg">d</a></td>
<td class="spare"></td>
</tr>
</tbody>
</table>
</div>
</p>
</div>
</div>
<div class="slide">
<H3>Inside marker: Final</H3>
<div class="H3">
<p>
<p>
In order to have just one object (i.e., connected component) inside each key, a dilation is applied.
</p>
<div class="example">
<div class="listing">
<pre class="user">>>> e = mmdil(d, mmsebox(3));</pre>
<pre class="computer"></pre>
<pre class="user">>>> mmshow(e);</pre>
<pre class="computer"></pre>
</div>
<table class="images">
<tbody align="center">
<tr class="image" valign="bottom">
<td><img width="295" src="../images/img_mmdcalc_005.jpg"></td>
<td class="spare"></td>
</tr>
<tr class="title" valign="baseline">
<td><a href="../images/img_mmdcalc_005.jpg">e</a></td>
<td class="spare"></td>
</tr>
</tbody>
</table>
</div>
</p>
</div>
</div>
<div class="slide">
<H3>Outside markers</H3>
<div class="H3">
<p>
<p>
The outside markers are built by taking the watershed (skiz) of the complement of internal markers image.
</p>
<div class="example">
<div class="listing">
<pre class="user">>>> f = mmwatershed(mmneg(e));</pre>
<pre class="computer"></pre>
<pre class="user">>>> mmshow(f);</pre>
<pre class="computer"></pre>
</div>
<table class="images">
<tbody align="center">
<tr class="image" valign="bottom">
<td><img width="295" src="../images/img_mmdcalc_006.jpg"></td>
<td class="spare"></td>
</tr>
<tr class="title" valign="baseline">
<td><a href="../images/img_mmdcalc_006.jpg">f</a></td>
<td class="spare"></td>
</tr>
</tbody>
</table>
</div>
</p>
</div>
</div>
<div class="slide">
<H3>Both markers</H3>
<div class="H3">
<p>
<p>
The markers used are the union of the internal and external markers detected. They are displayed as overlay on the gradient image.
</p>
<div class="example">
<div class="listing">
<pre class="user">>>> g = mmunion(e,f);</pre>
<pre class="computer"></pre>
<pre class="user">>>> mmshow(b,g);</pre>
<pre class="computer"></pre>
</div>
<table class="images">
<tbody align="center">
<tr class="image" valign="bottom">
<td><img width="295" src="../images/img_mmdcalc_007.jpg"></td>
<td class="spare"></td>
</tr>
<tr class="title" valign="baseline">
<td><a href="../images/img_mmdcalc_007.jpg">b,g</a></td>
<td class="spare"></td>
</tr>
</tbody>
</table>
</div>
</p>
</div>
</div>
<div class="slide">
<H3>Constrained watershed</H3>
<div class="H3">
<p>
<p>
The calculator keys are extracted by applying the watershed operator on the gradient image, constrained by the markers detected.
</p>
<div class="example">
<div class="listing">
<pre class="user">>>> h = mmcwatershed(b,g,mmsebox());</pre>
<pre class="computer"></pre>
<pre class="user">>>> mmshow(h);</pre>
<pre class="computer"></pre>
</div>
<table class="images">
<tbody align="center">
<tr class="image" valign="bottom">
<td><img width="295" src="../images/img_mmdcalc_008.jpg"></td>
<td class="spare"></td>
</tr>
<tr class="title" valign="baseline">
<td><a href="../images/img_mmdcalc_008.jpg">h</a></td>
<td class="spare"></td>
</tr>
</tbody>
</table>
</div>
</p>
</div>
</div>
<div class="slide">
<H3>Show the result</H3>
<div class="H3">
<p>
<p>
Superposition of the detected contours on the input image.
</p>
<div class="example">
<div class="listing">
<pre class="user">>>> mmshow(a,h);</pre>
<pre class="computer"></pre>
</div>
<table class="images">
<tbody align="center">
<tr class="image" valign="bottom">
<td><img width="295" src="../images/img_mmdcalc_009.jpg"></td>
<td class="spare"></td>
</tr>
<tr class="title" valign="baseline">
<td><a href="../images/img_mmdcalc_009.jpg">a,h</a></td>
<td class="spare"></td>
</tr>
</tbody>
</table>
</div>
</p>
</div>
</div>
</div>
</div>
<center>
<table class="botNav">
<tr>
<td class="index">
[<a href="../mmdemos/mmdbrain.html"><tt>mmdbrain</tt></a>]
[<a href="index.html">Up</a>]
[<a href="../mmdemos/mmdcells.html"><tt>mmdcells</tt></a>]
</td>
<td rowspan="2" class="xhtml"><a href="http://www.python.org"><img width="55" alt="Python" height="22" src="../PythonPoweredSmall.gif"></a></td>
</tr>
<tr>
<td class="copyright">Copyright (c) 2003, Roberto A. Lotufo, UNICAMP-University of Campinas; Rubens C. Machado, CenPRA-Renato Archer Research Center.</td>
</tr>
</table>
</center>
</body>
</html>
distrib
>
Mandriva
>
2010.0
>
i586
>
media
>
contrib-release
>
by-pkgid
>
4a71d9984febeb5a206904a5a379841a
>
files
>
615
