Re: best way to select a range of colors

Posted by Jacqueline Ross on Feb 10, 2015; 4:30am
URL: http://imagej.273.s1.nabble.com/best-way-to-select-a-range-of-colors-tp5011526p5011541.html

Hi Rodrigo,

Not sure exactly what you want to do with your analysis, assume it's counting particles, size, shape, etc.

I tried the Threshold Colour plugin on your image but that didn't work very well for me as your hues are not very well defined.

So, I went to my favourite, which is the Colour Deconvolution plugin. This is designed (by Gabriel Landini) for histological stains and I thought I'd test it out for your image. I found it worked pretty well after some colour balancing, background subtraction (radius 100) and using the ROI option, making sure to only draw an ROI around the really pink pixels and then selecting the brownish ones for the second colour.

You can see the result in the attached JPG montage. If you want the full processed images, please email me directly as I don't want to clog up the list.

The first image in the panel is the result of the colour balance and background subtraction. I then used that for the Colour Deconvolution.

Cheers,

Jacqui

Jacqueline Ross
Biomedical Imaging Microscopist
Biomedical Imaging Research Unit 
School of Medical Sciences 
Faculty of Medical & Health Sciences
The University of Auckland
Private Bag 92019
Auckland 1142, NEW ZEALAND

Tel: [phone]
Fax: [phone]

http://www.fmhs.auckland.ac.nz/sms/biru/


-----Original Message-----
From: ImageJ Interest Group [mailto:[hidden email]] On Behalf Of Rodrigo Gonçalves
Sent: Tuesday, 10 February 2015 3:38 a.m.
To: [hidden email]
Subject: best way to select a range of colors

Hi people, in the past I've used ImageJ/Fiji mostly for grayscale images, so I have little experience in the color world. Now I have some samples treated with a stain and I thought I would separate stained vs unstained particles. But it was not as straightforward as I expected (at least with the approach I'm using).

Here's the original image.
https://www.dropbox.com/s/yc7v90wv9s6rwje/initial_image.tif?dl=0

This is after I automatically re balance colors (for this step I used "auto adjust colors" in IrfanView, for some reason I can't get ImageJ to reach the same nice result with any of the "auto" options of the Color Balance tool):https://www.dropbox.com/s/lcmo93oce8rupze/step2_colors_autobalanced.tif?dl=0

In this image I pointed most of the interest particles so you can see the range of colors I'm looking for. https://www.dropbox.com/s/c78z37jbzsae223/some_interest_particles.jpg?dl=0

So the (unsuccessful) approach I tried is:1- manually select a few interest particles2- generate the color (RGB) histogram of those3- note the mean and standard deviation of each RGB4- with the Color Threshold tool, select the mean +/- 1-2 SD of each R, G and B

but this does not produce the desired threshold.
I'm sure this problem is as old as ImageJ so there must be tens of better ways, could you advice on this?Thanks a lot, Rodrigo
 ________________________
Rodrigo J. Gonçalves

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

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