Scale on images and their impact on measurements..Very Urgent. PLEASE AND THANK YOU

[Geology Guy]

Good Morning Everyone,
I have some Back Scattered electron images with particles that I want to measure. All images have a resolution of 2048 by 1536. Some of them have a scale of 2 microns, some 1 micron and others 3 microns. I have decided that I want to make my measurements in nanometers so I used a straight line tool to measure the scales on each micrograph and entered the unit using ANAYZE > SET Measurements. My question now is that since all of them are now in units of nanometers, do I still have to worry about the fact that they all had different scales i.e. 1, 2 or 3 microns? I am not sure if the fact that I had different scales to begin with means that I cannot compare the values of these images in terms of making scatter plots, bar charts, histograms and so on.

Thank you for responding on time

[Divakar Ramachandran-2]

I assume that you mean that the images, to start with, had different scale
markers of 1, 2 and 3 microns and that you have used analyse > set scale to
correlate the length of the scale markers to 1000, 2000 and 3000 nm, as
applicable for each image. Having done this, you could embed the new scale
information in to the image by saving them in TIFF format (AFAIK, this is
the only format that preserves the stored scale information in a way that
can be re-read when opened in ImageJ). Any subsequent quantification using
ImageJ should be as per this scale information, and would be independent of
the magnification at which the different images were recorded.

HTH,
Divakar


-----Original Message-----
From: Geology Guy
Sent: Wednesday, July 31, 2013 7:41 PM
To: [hidden email]
Subject: Scale on images and their impact on measurements..Very Urgent.
PLEASE AND THANK YOU

Good Morning Everyone,
I have some Back Scattered electron images with particles that I want to
measure. All images have a resolution of 2048 by 1536. Some of them have a
scale of 2 microns, some 1 micron and others 3 microns. I have decided that
I want to make my measurements in nanometers so I used a straight line tool
to measure the scales on each micrograph and entered the unit using ANAYZE >
SET Measurements. My question now is that since all of them are now in units
of nanometers, do I still have to worry about the fact that they all had
different scales i.e. 1, 2 or 3 microns? I am not sure if the fact that I
had different scales to begin with means that I cannot compare the values of
these images in terms of making scatter plots, bar charts, histograms and so
on.

Thank you for responding on time



--
View this message in context:
http://imagej.1557.x6.nabble.com/Scale-on-images-and-their-impact-on-measurements-Very-Urgent-PLEASE-AND-THANK-YOU-tp5004233.html
Sent from the ImageJ mailing list archive at Nabble.com.

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

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

[karo03]

Hi Geoplogy Guy,

today is July,31 perhaps not the date for making tests!!

However, asking is one way, trying another. What about setting the scale to two different values, e.g. a factor of two and perform the measurement? Length features like Major and Minor should directly reflect the factor, Area of course the squared factor, ratios (AR) of length should be independent of the factor and so on …

Regards

Karsten

Am 31.07.2013 um 16:11 schrieb Geology Guy <[hidden email]>:

> Good Morning Everyone,
> I have some Back Scattered electron images with particles that I want to
> measure. All images have a resolution of 2048 by 1536. Some of them have a
> scale of 2 microns, some 1 micron and others 3 microns. I have decided that
> I want to make my measurements in nanometers so I used a straight line tool
> to measure the scales on each micrograph and entered the unit using ANAYZE >
> SET Measurements. My question now is that since all of them are now in units
> of nanometers, do I still have to worry about the fact that they all had
> different scales i.e. 1, 2 or 3 microns? I am not sure if the fact that I
> had different scales to begin with means that I cannot compare the values of
> these images in terms of making scatter plots, bar charts, histograms and so
> on.
>
> Thank you for responding on time
>
>
>
> --
> View this message in context: http://imagej.1557.x6.nabble.com/Scale-on-images-and-their-impact-on-measurements-Very-Urgent-PLEASE-AND-THANK-YOU-tp5004233.html
> Sent from the ImageJ mailing list archive at Nabble.com.
>
> --
> ImageJ mailing list: http://imagej.nih.gov/ij/list.html

Karsten
[hidden email]

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

[Geology Guy]

Dear Divakar,

>I assume that you mean that the images, to start with, had different scale
>markers of 1, 2 and 3 microns and that you have used analyse > set scale to
>correlate the length of the scale markers to 1000, 2000 and 3000 nm, as
>applicable for each image............Any subsequent quantification using
>ImageJ should be as per this scale information, and would be independent of
>the magnification at which the different images were recorded.

Thank you very much for your reply and that is exactly what I meant and trying to achieve. I just wanted to make sure that once all the units are adjusted to nanometers I do not have to worry about what the original scale of the image was (i.e. 1, 2 and 3 microns) when I am making comparisons of the micrographs on a statistical plot. Thank very much once again for your help

[Geology Guy]

Hello Karsten,
Thank you very much for your reply and I think I am clear now on what needs to be done. Thank you

[Heeschen, Bill (WA)]

Hi G.G.
[This is a late - I use digest mode]

I think that from what you are saying, your measurements will all be correct in terms of calibrated lengths, etc.  This is consistent with other responeses I saw.  So, if all you're doing is comparing sizes, etc., your data are good, but...
Disclaimer:
This is assuming that your measurements are not influenced strongly by pixel size.  That is, if your object is 500 pixels long in one image and it would be 700 pixels long at a slightly different magnification, the error due to being off by one pixel is pretty much the same in both images.  (Errors of 1 part in 500 vs 1 part in 700 - nominally 0.2% - are typically overwhelmed by other sources of error in the method.)  Consider the other extreme where the object would be 5 pixels in one image and 7 in the other.  Now being off by one pixel generates errors more like 20% of the measured value in the 5 pixel case and 14% in the other.

There are a couple things that help with this +/- pixel error:
1)  As long as the error is random (as many times + as -) and you measure "many" particles (several hundred), the error will balance out in terms of the mean value, even for the 20% and 14% cases.  Note that the breadth of the distribution WILL be strongly affected at the 20% and 14% error (bad news), but at the 0.2% error range, other error sources will dominate any artificial spread of the distribution.
2)  Hopefully you have set your imaging conditions so that most of the particles of interest are closer to 500 pixels than 5 pixels.  In this case the overall error in breadth of the distribution due to pixel resolution will be tolerable since you're not particularly interested in the small ones, anyhow.
 
Let's assume you have plenty of pixels in all your images...


The challenge for dealing with different magnifications is if you are trying to compare some kind of concentration for a given size.  In that case you need to consider that the area of each image is not necessarily the same so that larger fields of view should naturally give you more counts of a given size.  You handle this by normalizing your counts to the area of the image.  For example, if you have two images and one of them is 2 microns x 2 microns (area = 4 micron^2)  and the other is 3 x 3 (area = 9 micron^2), then express all your cumulative measurements as value-per-square micron.

Number example:
Suppose I counted 36 objects that were 300 nm long in the 4 micron^2 image.  My normalized counts are then 9 objects per micron^2.
Now I count the 300 nm objects in the 9 micron^2 image and find 54.  This gives me a normalized count of 6 objects per micron^2.
The conclusion is that my concentration of 300 nm objects is lower in the "3 x 3" sample than in the "2 x 2" sample.

I hope this addresses the issue you raised!
Best Regards,


Bill
William A. Heeschen, Ph.D.
Microscopy, Digital Imaging
The Dow Chemical Company
Midland, MI  48667
[hidden email]

> Date:    Wed, 31 Jul 2013 07:11:44 -0700
> From:    Geology Guy <[hidden email]>
> Subject: Scale on images and their impact on measurements..Very Urgent. PLEASE AND THANK YOU
>
> Good Morning Everyone,
> I have some Back Scattered electron images with particles that I want to
> measure. All images have a resolution of 2048 by 1536. Some of them have a
> scale of 2 microns, some 1 micron and others 3 microns. I have decided that
> I want to make my measurements in nanometers so I used a straight line tool
> to measure the scales on each micrograph and entered the unit using ANAYZE >
> SET Measurements. My question now is that since all of them are now in units
> of nanometers, do I still have to worry about the fact that they all had
> different scales i.e. 1, 2 or 3 microns? I am not sure if the fact that I
> had different scales to begin with means that I cannot compare the values of
> these images in terms of making scatter plots, bar charts, histograms and so
> on.

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