processing of noisy images with particles in bad contrast

[M S]

Good day, dear ImageJ users!
My field of work is totally not related to image processing, but therealways is a first time when you need to do something new. Having no idea how toproceed makes it difficult though.
I have acquired many images similar to what can be seen in particles.tif. Ifyou use your imagination then you can see approximately 58 black blobs which Ihave marked in markers_particles.tif. The problem is that very low exposition wasused (since the particles move really fast) and that creates all the noisewhere actual particles are drowning.
I have played with ImageJ/Fiji settings – Image arithmetics, Threshold,Enhance Contrast, Noise removal etc. in different combinations, but nothingbrings any satisfactory results. By satisfactory I mean more or less clearly visibleparticles that could be traced. If from these 50-something particles that canbe seen even half could be traced, that would be a great achievement. I haveattached a short preview of image sequence as well.
So the question – is it possible to automatically process these imagesequences in not time consuming way (manual processing of single image is notwhere I would like to spend my time) so final result would be a clear view ofat least some particles which could be traced using for example Mosaic’sParticle Tracker?
Sincerely,John

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Jerry (Gerald) Sedgewick]

A deceptively easy way to improve contrast and remove noise can be done
through resampling. If you scale (resample) your image to a less pixels
across and down (say 25% of its original pixel resolution) then resample
again to enlarge it back to its original size, you'll see the black spots
sans noise (if these are the ones that are larger in your image). Experiment
with different resampling percentages to get the desired result.

Jerry

-----Original Message-----
From: M S
Sent: Friday, July 31, 2015 7:30 AM
To: [hidden email]
Subject: processing of noisy images with particles in bad contrast

Good day, dear ImageJ users!
My field of work is totally not related to image processing, but therealways
is a first time when you need to do something new. Having no idea how
toproceed makes it difficult though.
I have acquired many images similar to what can be seen in particles.tif.
Ifyou use your imagination then you can see approximately 58 black blobs
which Ihave marked in markers_particles.tif. The problem is that very low
exposition wasused (since the particles move really fast) and that creates
all the noisewhere actual particles are drowning.
I have played with ImageJ/Fiji settings – Image arithmetics,
Threshold,Enhance Contrast, Noise removal etc. in different combinations,
but nothingbrings any satisfactory results. By satisfactory I mean more or
less clearly visibleparticles that could be traced. If from these
50-something particles that canbe seen even half could be traced, that would
be a great achievement. I haveattached a short preview of image sequence as
well.
So the question – is it possible to automatically process these
imagesequences in not time consuming way (manual processing of single image
is notwhere I would like to spend my time) so final result would be a clear
view ofat least some particles which could be traced using for example
Mosaic’sParticle Tracker?
Sincerely,John

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html 

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Michael Schmid]

Hi John,

it's so noisy that it is difficult to decide what is a blob and what isn't.
For processing, first you have to get rid of the margins with their different contrast. Then some combinations of filtering and smoothing will give at least some reasonable result. This is what I tried:

makeRectangle(6, 22, 487, 470);
run("Crop");
setThreshold(254, 255);
run("Create Selection");
run("Smooth");
run("Select None");
run("Smooth");
run("Remove Outliers...", "radius=3 threshold=20 which=Bright");
run("Gaussian Blur...", "sigma=2");
run("Maximum...", "radius=1");
run("Subtract Background...", "rolling=10 light");
setAutoThreshold("MaxEntropy");

Then run 'Analyze Particles' with a minimum size of your choice, depending on what minimum size you consider a true blob, not just noise.

Michael
________________________________________________________________
On Jul 31, 2015, at 14:30, M S wrote:

> Good day, dear ImageJ users!
> My field of work is totally not related to image processing, but therealways is a first time when you need to do something new. Having no idea how toproceed makes it difficult though.
> I have acquired many images similar to what can be seen in particles.tif. Ifyou use your imagination then you can see approximately 58 black blobs which Ihave marked in markers_particles.tif. The problem is that very low exposition wasused (since the particles move really fast) and that creates all the noisewhere actual particles are drowning.
> I have played with ImageJ/Fiji settings – Image arithmetics, Threshold,Enhance Contrast, Noise removal etc. in different combinations, but nothingbrings any satisfactory results. By satisfactory I mean more or less clearly visibleparticles that could be traced. If from these 50-something particles that canbe seen even half could be traced, that would be a great achievement. I haveattached a short preview of image sequence as well.
> So the question – is it possible to automatically process these imagesequences in not time consuming way (manual processing of single image is notwhere I would like to spend my time) so final result would be a clear view ofat least some particles which could be traced using for example Mosaic’sParticle Tracker?
> Sincerely,John
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html<particles_anim.tif><markers_particles.tif><particles.tif>

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html

[Jean-Philippe Grossier]

Hi,

I tried to subtract two gaussian with a small and a big radius and it gives
decent results:

makeRectangle(13, 41, 470, 445);
run("Crop");

run("Duplicate...", "title=bigGauss duplicate");
run("Invert", "stack");
run("Gaussian Blur...", "sigma=15 stack");


selectWindow("particles_anim.tif");
run("Duplicate...", "title=smallGauss duplicate");
run("Invert", "stack");
run("Gaussian Blur...", "sigma=5 stack");

imageCalculator("Subtract create stack", "smallGauss","bigGauss");
selectWindow("Result of smallGauss");
setMinAndMax(9, 11);



you can play with the radius to get better result I guess.

Jean-Philippe

2015-07-31 9:10 GMT-04:00 J. Sedgewick <[hidden email]>:

> A deceptively easy way to improve contrast and remove noise can be done
> through resampling. If you scale (resample) your image to a less pixels
> across and down (say 25% of its original pixel resolution) then resample
> again to enlarge it back to its original size, you'll see the black spots
> sans noise (if these are the ones that are larger in your image).
> Experiment with different resampling percentages to get the desired result.
>
> Jerry
>
> -----Original Message----- From: M S
> Sent: Friday, July 31, 2015 7:30 AM
> To: [hidden email]
> Subject: processing of noisy images with particles in bad contrast
>
>
> Good day, dear ImageJ users!
> My field of work is totally not related to image processing, but
> therealways is a first time when you need to do something new. Having no
> idea how toproceed makes it difficult though.
> I have acquired many images similar to what can be seen in particles.tif.
> Ifyou use your imagination then you can see approximately 58 black blobs
> which Ihave marked in markers_particles.tif. The problem is that very low
> exposition wasused (since the particles move really fast) and that creates
> all the noisewhere actual particles are drowning.
> I have played with ImageJ/Fiji settings – Image arithmetics,
> Threshold,Enhance Contrast, Noise removal etc. in different combinations,
> but nothingbrings any satisfactory results. By satisfactory I mean more or
> less clearly visibleparticles that could be traced. If from these
> 50-something particles that canbe seen even half could be traced, that
> would be a great achievement. I haveattached a short preview of image
> sequence as well.
> So the question – is it possible to automatically process these
> imagesequences in not time consuming way (manual processing of single image
> is notwhere I would like to spend my time) so final result would be a clear
> view ofat least some particles which could be traced using for example
> Mosaic’sParticle Tracker?
> Sincerely,John
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html
>

--
ImageJ mailing list: http://imagej.nih.gov/ij/list.html