Resend from bounce...
On Fri, Apr 24, 2015 at 3:05 PM, Chris Marshall <[email protected]>
wrote:
> Nice with the IPython notebooks, Zaki.
>
> Here is a sample code showing processingusing color in addition to
> intensity.
> In this type of processing, the real workcomes in cleaning up the data,
> removing
> flaws and irregularities,...
>
> # Color segmentation and processing for chromatography
>> # An example of using PDL for image processing
>> use PDL;
>> use PDL::NiceSlice;
>> use PDL::IO::Pic;
>> use Time::HiRes qw(sleep);
>> use PDL::Graphics2D;
>> use PDL::Image2D;
>>
>> # Read in image data, mv() used to put pixel axis in dim(0)
>> $tlc = rim('TLC_arg.jpg')->mv(-1,0);
>> imag2d $tlc, 'TLC', 0.25;
>>
>> # Select background pixels based on the white part of the
>> # image and calculate the average color...
>> #
>> # Convert RGB to greyscale by calculated the average pixel value
>> $tlc_g = $tlc->float->average;
>>
>> # Background is the top 50% pixels
>> $back_mask = ($tlc_g > $tlc_g->pct(0.5));
>> imag2d $back_mask, 'background mask', 0.25;
>>
>> # Average background RGB value
>> $background = $tlc->mv(0,-1)->whereND($back_mask);
>> $b_rgb = $background->float->average;
>>
>> # Compute color direction vector
>> $b_color = $b_rgb->norm;
>>
>> # Calculate the color direction for each pixel
>> $tlc_unit = $tlc->float->norm;
>> imag2d $tlc_unit, 'TLC unit', 0.25;
>>
>> # We calculate the deviation from the background direction
>> $tlc_cos = $tlc_unit->inner($b_color);
>>
>> # Switch to sin which is proportional to angle
>> $tlc_sin = sqrt(1-$tlc_cos);
>> $tlc_sin /= $tlc_sin->max;
>>
>> # Background is at 0 and color differences are positive
>> imag2d $tlc_sin, 'tlc color angle', 0.25;
>>
>> # Look at hist values to pick color delta threshhold to make mask
>> print cat($tlc_sin->hist(0,1,0.05))->mv(0,1), "\n";
>> $spot_mask = ($tlc_sin>0.1); # 10% looks good
>> imag2d $spot_mask->float, 'spot mask', 0.25;
>>
>> # Calculate connected clusters in the mask image
>> $cc8 = $spot_mask->float->cc8compt;
>> imag2d $cc8/$cc8->max, 'cc8 big', 0.25;
>>
>> # Calculate cluster sizes from segement index values
>> ($num,$label) = $cc8->flat->qsort->rle;
>> $maxind = which($num>0)->max;
>>
>> # Chop off the trailing zeros from rle result
>> $num = $num(0:$maxind);
>> $label = $label(0:$maxind);
>>
>> # Clean up clusters by removing below a certain size
>> # Take a look at the sizes to find a cut point
>> print cat($num,$label)->mv(0,1)->qsortvec, "\n";
>>
>> # loop over cluster values and clear if size < 100
>> foreach my $ind ( $label->where($num < 100)->list ) { $cc8->where($cc8
>> == $ind) .= 0 };
>>
>> # More segementation could use hue variation in addition to difference
>> from background
>> $tlc_cross = $tlc_unit->crossp($b_color);
>> $xxx = $tlc_cross / ($tlc_cross->stats)[1];
>> $xxx = $xxx->rint->clip(-1,1);
>> imag2d $xxx/2+0.5, 'binary box', 0.25;
>>
>> # look at which pixels correspond to the uniq triplet values from $xxx
>> # such a [1,0,0], [1,1,0], ...
>> $plane = $xxx->norm->(:,*1)->inner($xxx->uniqvec->norm);
>> $plane->where($plane < 0.9999) .= 0;
>> $plane->where($plane >= 0.9999) .= 1;
>>
>> # loop and show for now
>> $win = imag2d $plane((0)), 'plane', 0.25;
>> $loopmod = $plane->dim(0);
>> foreach my $level ( 0..1000 ) { my $ind = $level%$loopmod; imag2d_update
>> $win, $plane(($ind)); sleep 0.5; }
>>
>>
> --Chris
>
> On Thu, Apr 23, 2015 at 4:16 PM, Demian Riccardi <[email protected]
> > wrote:
>
>> holy smokes!
>>
>> > On Apr 23, 2015, at 4:12 PM, Zakariyya Mughal <[email protected]>
>> wrote:
>> >
>> > On 2015-04-23 at 11:06:38 -0600, Derek Lamb wrote:
>> >>
>> >>> On Apr 23, 2015, at 9:34 AM, Zakariyya Mughal <[email protected]>
>> wrote:
>> >>>
>> >>> Ignore my e-mail about using clustering algorithms... Craig's e-mail
>> >>> reminded my that you can do blob extraction in one-pass using
>> >>> <
>> https://en.wikipedia.org/wiki/Connected-component_labeling#One-pass_version
>> >.
>> >>> There's some C code at <
>> http://www-sop.inria.fr/members/Gregoire.Malandain/segment/connexe.html>,
>> >>> but I haven't used it.
>> >>>
>> >>> Cheer,
>> >>> - Zaki Mughal
>> >>
>> >> Even better, PDL already has connected-componenet labeling! See
>> cc8compt, cc4compt, and ccNcompt.
>> >>
>> >> http://pdl.perl.org/?docs=Image2D&title=PDL::Image2D#cc8compt
>> >
>> > Thanks!
>> >
>> > I gave it a go at <
>> http://nbviewer.ipython.org/gist/zmughal/fae2a5901d59504b56e1/Thin-layer-chromatography-analysis.ipynb
>> >.
>> >
>> > I'm going to have to compare this with the MATLAB equivalent at some
>> > point to see if PDL's version can be made simpler.
>> >
>> > Cheers,
>> > - Zaki Mughal
>> >
>> >>
>> >> Derek
>> >
>> >
>> ------------------------------------------------------------------------------
>> > BPM Camp - Free Virtual Workshop May 6th at 10am PDT/1PM EDT
>> > Develop your own process in accordance with the BPMN 2 standard
>> > Learn Process modeling best practices with Bonita BPM through live
>> exercises
>> > http://www.bonitasoft.com/be-part-of-it/events/bpm-camp-virtual-
>> event?utm_
>> > source=Sourceforge_BPM_Camp_5_6_15&utm_medium=email&utm_campaign=VA_SF
>> > _______________________________________________
>> > pdl-general mailing list
>> > [email protected]
>> > https://lists.sourceforge.net/lists/listinfo/pdl-general
>>
>>
>>
>> ------------------------------------------------------------------------------
>> BPM Camp - Free Virtual Workshop May 6th at 10am PDT/1PM EDT
>> Develop your own process in accordance with the BPMN 2 standard
>> Learn Process modeling best practices with Bonita BPM through live
>> exercises
>> http://www.bonitasoft.com/be-part-of-it/events/bpm-camp-virtual-
>> event?utm_
>> source=Sourceforge_BPM_Camp_5_6_15&utm_medium=email&utm_campaign=VA_SF
>> _______________________________________________
>> pdl-general mailing list
>> [email protected]
>> https://lists.sourceforge.net/lists/listinfo/pdl-general
>>
>
>
------------------------------------------------------------------------------
One dashboard for servers and applications across Physical-Virtual-Cloud
Widest out-of-the-box monitoring support with 50+ applications
Performance metrics, stats and reports that give you Actionable Insights
Deep dive visibility with transaction tracing using APM Insight.
http://ad.doubleclick.net/ddm/clk/290420510;117567292;y
_______________________________________________
pdl-general mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/pdl-general
