Great!<div><br></div><div>Now that I look at this, I realise that the comparison is not fair to mahotas because it includes a thresholding step. The correct comparison would run mahotas.sobel(f, just_filter=True) which runs in almost the exact same time as skimage because it is almost the exact same code ;-)</div><div><br></div><div>mahotas.center_of_mass also handles labeled images; btw.</div><div><br></div><div>Best,</div><div>Luis<br><br>On Tuesday, September 4, 2012 4:22:54 PM UTC+1, Johannes Schönberger wrote:<blockquote class="gmail_quote" style="margin: 0;margin-left: 0.8ex;border-left: 1px #ccc solid;padding-left: 1ex;">After another round of performance improvements for the sobel filter, I get:<div><br></div><blockquote style="margin:0 0 0 40px;border:none;padding:0px"><div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left-width:1px;border-left-color:rgb(204,204,204);border-left-style:solid;padding-left:1ex">erode &nbsp; &nbsp; &nbsp; &amp; &nbsp; &nbsp; 9.03 &amp; &nbsp; &nbsp;13.67 &amp; &nbsp; &nbsp;13.31 &amp; \\<br>dilate &nbsp; &nbsp; &nbsp;&amp; &nbsp; &nbsp;12.45 &amp; &nbsp; &nbsp; 9.40 &amp; &nbsp; &nbsp; 6.60 &amp; \\<br>open &nbsp; &nbsp; &nbsp; &nbsp;&amp; &nbsp; &nbsp;21.99 &amp; &nbsp; &nbsp;22.70 &amp; &nbsp; &nbsp;13.25 &amp; \\<br>center mass &amp; &nbsp; &nbsp; 6.63 &amp; &nbsp; &nbsp; &nbsp; NA &amp; &nbsp; &nbsp; &nbsp; NA &amp; \\<br>sobel &nbsp; &nbsp; &nbsp; &amp; &nbsp; &nbsp;74.66 &amp; &nbsp; &nbsp; &nbsp; NA &amp; &nbsp; &nbsp;43.72 &amp; \\</blockquote></div></blockquote><div><br></div><div>Note, that there also exists a possibility to calculate the center of mass using the `regionprops` function and the properties "Centroid" and "WeightedCentroid". But due to the label overhead this is probably a lot slower than your implementation.</div><div><div><br></div><br>Am Donnerstag, 30. August 2012 18:52:59 UTC+2 schrieb Luis Pedro Coelho:<blockquote class="gmail_quote" style="margin:0;margin-left:0.8ex;border-left:1px #ccc solid;padding-left:1ex">Hello,
<br>
<br>In the context of my previous email to pythonvision
<br>
<br><a href="https://groups.google.com/d/topic/pythonvision/SSCF4LUlRKQ/discussion" target="_blank">https://groups.google.com/d/<wbr>topic/pythonvision/<wbr>SSCF4LUlRKQ/discussion</a>
<br>
<br>I made a timing comparison between mahotas &amp; scikits-image.
<br>
<br>operation &nbsp; | mahotas &nbsp;| pymorph &nbsp;| skimage
<br>------------+----------+------<wbr>----+---------
<br>erode &nbsp; &nbsp; &nbsp; | &nbsp; &nbsp;10.80 | &nbsp; &nbsp;14.33 | &nbsp; &nbsp;80.17
<br>dilate &nbsp; &nbsp; &nbsp;| &nbsp; &nbsp;11.44 | &nbsp; &nbsp; 8.93 | &nbsp; &nbsp;41.59
<br>open &nbsp; &nbsp; &nbsp; &nbsp;| &nbsp; &nbsp;22.45 | &nbsp; &nbsp;23.20 | &nbsp; &nbsp;80.18
<br>center mass | &nbsp; &nbsp; 7.05 | &nbsp; &nbsp; &nbsp; NA | &nbsp; &nbsp; &nbsp; NA
<br>sobel &nbsp; &nbsp; &nbsp; | &nbsp; &nbsp;75.03 | &nbsp; &nbsp; &nbsp; NA | &nbsp; 105.72
<br>cwatershed &nbsp;| &nbsp; 201.03 | 56586.50 | &nbsp; 290.41
<br>daubechies &nbsp;| &nbsp; &nbsp;19.05 | &nbsp; &nbsp; &nbsp; NA | &nbsp; &nbsp; &nbsp; NA
<br>haralick &nbsp; &nbsp;| &nbsp; 306.48 | &nbsp; &nbsp; &nbsp; NA | &nbsp;7391.37
<br>
<br>(Best viewed with fixed-width fonts)
<br>
<br>The unit of measurement is the time it takes to run ``numpy.max(image)`` 
<br>Mahotas is always faster than skimage (although pymorph is better for certain 
<br>morphological basic operations). I used GCLM in skimage to stand for Haralick, 
<br>which is a rough approximation.
<br>
<br>I attach the script that generates these (against github skimage). If you 
<br>think that I have used skimage incorrectly, please let me know.
<br>
<br>Best,
<br>-- 
<br>Luis Pedro Coelho | Institute for Molecular Medicine | <a href="http://luispedro.org" target="_blank">http://luispedro.org</a></blockquote></div></blockquote></div>