Assessing bird colouration

[Ross Ahmed]

The following paper details how a spectrometer was used to analyse the
colouration of a type of bird called a Hoopoe:
http://www.nrm.se/download/18.382ec743113b0c5a2a180003662/Ericson+BOC+1997+Upupa.pdf

In the paper above, the spectrometer was used to analyse the colouration of
museum specimens of Hoopoe (i.e. the birds are dead/stuffed). Is it possible
to generate similar results by analysing a photo of
a dead/stuffed Hoopoe using Image J?

Any advice or comments welcome. Thanks in advance.

Ross

[Gabriel Landini]

On Sunday 19 December 2010 19:38:18 Ross Ahmed wrote:
> In the paper above, the spectrometer was used to analyse the colouration of
> museum specimens of Hoopoe (i.e. the birds are dead/stuffed). Is it possible
> to generate similar results by analysing a photo of a dead/stuffed Hoopoe
> using Image J?

Do you mean using an RGB image? No, you can't.
You need spectrophotometer or a multispectral setup (for example a camera with
a tunable filter or a collection of band pass filters).
Cheers
Gabriel

[Robert Dougherty]

Gabriel,

Ericson seems to have based his analysis on a 3-D color space "designed to fit the human perception of color hues."  Since the RGB / HSB color space used in ImageJ is also presumably oriented around human perception, is it possible that an RGB image might be sufficient in this particular case?  I'm suggesting that if the illumination and the RBG filters in the camera and everything else were lined up correctly, that it might be possible to get equivalent results.  Even if the color values did not exactly match Ericon's, they still might be rich enough to classify  Hoopoes by sex and subspecies.

Bob


On Dec 19, 2010, at 12:21 PM, Gabriel Landini wrote:

> On Sunday 19 December 2010 19:38:18 Ross Ahmed wrote:
>> In the paper above, the spectrometer was used to analyse the colouration of
>> museum specimens of Hoopoe (i.e. the birds are dead/stuffed). Is it possible
>> to generate similar results by analysing a photo of a dead/stuffed Hoopoe
>> using Image J?
>
> Do you mean using an RGB image? No, you can't.
> You need spectrophotometer or a multispectral setup (for example a camera with
> a tunable filter or a collection of band pass filters).
> Cheers
> Gabriel

Robert Dougherty, Ph.D.
President, OptiNav, Inc.
1414 127th Place NE #106
Bellevue, WA 98005
[phone]
FAX [phone]
www.optinav.com
[hidden email]

[Gary Sellani]

I think this would be possible IF you could process the raw image from the camera.

-----Original Message-----
From: Robert Dougherty <[hidden email]>
Sender: ImageJ Interest Group <[hidden email]>
Date: Sun, 19 Dec 2010 15:58:47
To: <[hidden email]>
Reply-To: ImageJ Interest Group <[hidden email]>
Subject: Re: Assessing bird colouration

Gabriel,

Ericson seems to have based his analysis on a 3-D color space "designed to fit the human perception of color hues."  Since the RGB / HSB color space used in ImageJ is also presumably oriented around human perception, is it possible that an RGB image might be sufficient in this particular case?  I'm suggesting that if the illumination and the RBG filters in the camera and everything else were lined up correctly, that it might be possible to get equivalent results.  Even if the color values did not exactly match Ericon's, they still might be rich enough to classify  Hoopoes by sex and subspecies.

Bob


On Dec 19, 2010, at 12:21 PM, Gabriel Landini wrote:

> On Sunday 19 December 2010 19:38:18 Ross Ahmed wrote:
>> In the paper above, the spectrometer was used to analyse the colouration of
>> museum specimens of Hoopoe (i.e. the birds are dead/stuffed). Is it possible
>> to generate similar results by analysing a photo of a dead/stuffed Hoopoe
>> using Image J?
>
> Do you mean using an RGB image? No, you can't.
> You need spectrophotometer or a multispectral setup (for example a camera with
> a tunable filter or a collection of band pass filters).
> Cheers
> Gabriel

Robert Dougherty, Ph.D.
President, OptiNav, Inc.
1414 127th Place NE #106
Bellevue, WA 98005
[phone]
FAX [phone]
www.optinav.com
[hidden email]

[dscho]

Hi,

On Sun, 19 Dec 2010, Robert Dougherty wrote:

> Ericson seems to have based his analysis on a 3-D color space "designed
> to fit the human perception of color hues."  Since the RGB / HSB color
> space used in ImageJ is also presumably oriented around human
> perception, is it possible that an RGB image might be sufficient in this
> particular case?

First of all, there are so many different definitions of RGB out there
that you cannot speak of a single RGB space.

> I'm suggesting that if the illumination and the RBG filters in the
> camera and everything else were lined up correctly, that it might be
> possible to get equivalent results.  Even if the color values did not
> exactly match Ericon's, they still might be rich enough to classify
> Hoopoes by sex and subspecies.

You can actually convert from sRGB (or any well-defined RGB) to CIELAB.
There is even a weakly-tested version of this algorithm (expecting the
current RGB image to be in sRGB, really) in Fiji's Image>Color menu.

Whether that is enough to reproduce the results of the paper remains to be
tested, though (the paper is really funny in that they go out of their way
to record a full-blown spectrogram, but then they reduce it to CIELAB
again...).

Ciao,
Johannes

[Wolfgang Schechinger]

If you can *see* the differences between the birds with your eyes in pictures taken with a camera that takes RGB images, I think, it should be possible. Maybe you need some sort of "standard curve" / reference data sets created by visual assertion of birds to groups.
Actually this might be a good job for a neuronal network.

Have fun

Wo
--
GRATIS! Movie-FLAT mit über 300 Videos.
Jetzt freischalten unter http://portal.gmx.net/de/go/maxdome

[dscho]

Hi,

On Mon, 20 Dec 2010, Gary Sellani wrote:

> I think this would be possible IF you could process the raw image from
> the camera.

That is only true iff the camera is color-calibrated exactly. You can find
out by using the paper color charts Gabriel mentioned. Even then I am not
sure if all RAW-capable cameras (I would certainly expect it from high-end
cameras, but only a control experiment will show!) correct for different
lighting conditions (brightness, light source) appropriately. As Michael
mentioned, this can make a crucial difference.

In short: if there is _any_ chance to use a spectrometer, you should use
it. These guys are not exactly cheap (really good ones are about $1,500 if
my research on that was correct), but as Gabriel suggests, there are
opportunities to collaborate with people who own one already
(psychologists interested in color perception, designers, material
scientists, etc).

Ciao,
Johannes