measuring darkness- greyscale calibration

[AmarantaFontcuberta]

Hello,

I'm trying to compare the wing darkness of many butterfly pictures.

For this puropose, I understood I can measure the mean grey value of a selected area within the wing.

The light conditions of the pictures may not be exactly the same, and for this reason I need to calibrate the grey scale for each picture.

Because I didn't know beforehand how to do it, I took pictures of my specimens with a color scale (that contains 5 basic colors + black+ white).

Can I calibrate the grey scale with this? (without having the nice grey bars scale)
In case not, is there any other mesure I can do with my color pictures to estimate darkness? How?


Thanks a lot for your help
:-)

Cheers,

Amaranta.

[Michael Schmid]

Hi Amarantha,

first a word of caution: A digital consumer camera is by no means a linear detector, so the following is a very inaccurate procedure. Also, you have to care about vignetting and uneven illumination (quite obvious in your case!).

What you could do as a rough approximation:

Convert the image to Composite
For each of the 3 color channels:
-  Measure the brightness of an area that is the same in all of your images (select the area, 'Measure' with 'Mean' selected in 'Measurement Options').  Due to the uneven illumination, I'd propose to use the blueish background near the objects of interests (wings) as a standard.
-  Multiply the image by a value that you calculate as constant_value/mean brightness. The 'constant_value' should be always the same, roughly the typical mean brightness of that area in the given color.

When necessary, convert to RGB again.

When finally measuring brightness of an object in RGB image (not the individual color channels), note that there are different ways: You have to determine the 'weight' of each color. For a rough match to, human perception, use 'Weighted RGB' in Edit>Options>Conversions.


Michael
________________________________________________________________
On May 10, 2012, at 09:59, AmarantaFontcuberta wrote:

> Hello,
>
> I'm trying to compare the wing darkness of many butterfly pictures.
>
> For this puropose, I understood I can measure the mean grey value of a
> selected area within the wing.
>
> The light conditions of the pictures may not be exactly the same, and for
> this reason I need to calibrate the grey scale for each picture.
>
> Because I didn't know beforehand how to do it, I took pictures of my
> specimens with a color scale (that contains 5 basic colors + black+ white).
>
> Can I calibrate the grey scale with this? (without having the nice grey bars
> scale)
> In case not, is there any other mesure I can do with my color pictures to
> estimate darkness? How?
>
>
> Thanks a lot for your help
> :-)
>
> Cheers,
>
> Amaranta.
>
> http://imagej.1557.n6.nabble.com/file/n4965116/05-DSC_0369.jpg 
>
> --
> View this message in context: http://imagej.1557.n6.nabble.com/measuring-darkness-greyscale-calibration-tp4965116.html
> Sent from the ImageJ mailing list archive at Nabble.com.

[AmarantaFontcuberta]

Hello Michael,

Thanks very much for your detailed answer.

Following your procedure I have some doubts.

-What do you mean for the constant_value as "roughly the typical mean brightness "? Do I guess it myself or is it calculated somewhere?
 Should it be the same between the three channels or between all pictures?

- Then I multiply the picture 3 times (the constant_value divided per each mean brightness), right?

-I go back to RGB and the picture now will be balanced. I set the weighted RGB. Select my area and measure  "mean grey value" again.

Thank you


Amaranta.

[Michael Schmid]

Hi Amaranta,

concerninng (1) constant_value as "roughly the typical mean brightness":
The idea is the following: say, your background typically has a brightness of 190 in R (red).  Then, you might multiply the intensity of all images to get a background of exactly 190.  If you choose a value that is too high, say, 255, then everything brighter than 190 will become saturated.  On the other hand, if you multiply the intensities by a low value, your accuracy will suffer from digitization noise (steps between the gray levels).

(2) You have to select such a constant for each of the 3 color channels.  In other words, you have to normalize the intensity of each channel individually.  A common constant for each channel would be fine only if you had a gray or white background, but that's not the case.

(3) go back to RGB, weighted RGB, and measure: Yes, that's what I wanted to say.


One more remark: you could try to measure the intensity of one of the dark color standards in your normalized images, to get an idea about the accuracy the procedure.

And finally, beware that the color of some butterflies can depend a lot on the illumination angle!  That's true especially for those where the color comes from interference of light at structures having dimensions in the range of light's wavelength.

Michael
________________________________________________________________
On May 14, 2012, at 10:47, AmarantaFontcuberta wrote:

> Hello Michael,
>
> Thanks very much for your detailed answer.
>
> Following your procedure I have some doubts.
>
> -What do you mean for the constant_value as "roughly the typical mean
> brightness "? Do I guess it myself or is it calculated somewhere?
> Should it be the same between the three channels or between all pictures?
>
> - Then I multiply the picture 3 times (the constant_value divided per each
> mean brightness), right?
>
> -I go back to RGB and the picture now will be balanced. I set the weighted
> RGB. Select my area and measure  "mean grey value" again.
>
> Thank you
>
>
> Amaranta.
>
> --
> View this message in context: http://imagej.1557.n6.nabble.com/measuring-darkness-greyscale-calibration-tp4965116p4975650.html
> Sent from the ImageJ mailing list archive at Nabble.com.

[AmarantaFontcuberta]

Hi!
Thanks again.. Now it seems much more clear.

The image multiplicator doesn't allow to multiply per a rational number, only entire numbers? No decimals?

The considerations about changing colors are completely true.

Thanks a lot


Amaranta.

[Michael Schmid]

Hi Amaranta,

Process>Math>Multiply allows you to have floating-point numbers as multipliers, but the result will be rounded to an integer value, 0-255, if you have an 8-bit image.

Michael
________________________________________________________________
On May 14, 2012, at 15:44, AmarantaFontcuberta wrote:

> Hi!
> Thanks again.. Now it seems much more clear.
>
> The image multiplicator doesn't allow to multiply per a rational number,
> only entire numbers? No decimals?
>
> The considerations about changing colors are completely true.
>
> Thanks a lot
>
>
> Amaranta.
>
> --
> View this message in context: http://imagej.1557.n6.nabble.com/measuring-darkness-greyscale-calibration-tp4965116p4975976.html
> Sent from the ImageJ mailing list archive at Nabble.com.