To help you test your solution, an offline tester is provided which will allow
you to test the examples.  As usual, to use it you should modify your program
to read from standard in and write to standard out.  Your program should read
the parameters one at a time, each on its own line.  You should then write
your output array, one element at a time, to standard out, with whitespace
delimiting elements (be sure to flush your output at the end).  <br/><br/>
A <a href="sample.html">sample</a> solution that works this way is provided.
You will notice that this is no different from the solution one would submit
on the website, except that a main function has been added.
<br/><br/>
To run the examples offline, you must download and extract the contents of <a
    href=ImageFilter.zip>ImageFilter.zip</a>.  When running your code, you
must specify the seed for the random number generator, the base filename
of the image files, and the executable command (using -seed, -base, and
-exec).  To run things, you must have the files <tt>&lt;base&gt;.raw</tt> and
<tt>&lt;base&gt;.result</tt> You may then specify one of a number of optional
parameters:

<tt>
<ul>
    <li>-compare    compares your output to the correct output</li>
    <li>-diff       compares your output to the correct output and computes the difference</li>
    <li>-view       view the result: either your output or the difference between the two if -diff is specified</li>
    <li>-output     generate &lt;base&gt;.you.png and &lt;base&gt;.diff.png (if -diff is specified)</li>
</ul>
</tt>
When rendering your result, it is scaled by taking <br/>
<tt>floor((intensity - min) / (max - min) * 255.999)</tt><br/>
If the correct result was loaded, the min and max values from it are used
(over all channels).  Otherwise, the min and max values of your image are
used.  The .raw.png and .result.png files provided are also
scaled in this way (using their own mins and maxes).  When rending the difference, it is scaled as:<br/>
<tt>floor(abs(diff)/max*255.999)</tt>.   All the images are rendered using
only 8 bits per channel.<br/><br/>
<ul>
    <li><a href="0.zip">Example 0</a></li>
    <li><a href="1.zip">Example 1</a></li>
    <li><a href="2.zip">Example 2</a></li>
    <li><a href="3.zip">Example 3</a></li>
    <li><a href="4.zip">Example 4</a></li>
    <li><a href="5.zip">Example 5</a></li>
    <li><a href="6.zip">Example 6</a></li>
    <li><a href="7.zip">Example 7</a></li>
    <li><a href="8.zip">Example 8</a></li>
    <li><a href="9.zip">Example 9</a></li>
</ul>
For example, after downloading <a href="ImageFilter.zip">ImageFilter.zip</a>
and <a href="0.zip">example 0</a>, you could compile the sample solution by
running:<br/>
<tt>g++ sol.cpp -o3 -o sol</tt><br/>
You could then run:<br/>
<tt>java -Xmx512M Runner -exec ./sol -seed 0 -base 0 -view -output
    -diff</tt><br/><br/>
"-seed 0" tells the program how to generate the filter (the seeds are the same
as the example numbers).  "-base 0" tells the program to look for 0.raw and
0.result for the input and correct output.  "-view" tells the program to
display your result on the screen, and "-output" tells it to save
"0.you.png".<br/><br/>
To generate your own tests, you will first have to extract the color values
from an image into a file with the same format as the ".raw" files (width, the
height, then RGB values in row major order as integers from 0 to 65535, where
the red, green and blue values of each pixel are grouped together -- in the
example files each line represents one pixel).<br/><br/>
Be aware of whether commas or periods are being used as decimal points.  If you want a decimal point, but are in a non-English locale, you can run:<br/>
<tt>java -Duser.language=en ...</tt>