Measuring color difference on spiders

[hkellogg]

Not sure where to post this but Color Management seemed the most logical.

I have a biologist friend that is doing research on the change of color of crab spiders measured from digital color images opened within Photoshop. I had suggested to use the RAW file format to eliminate the color conversion that would happen with the jpeg file format. He was interested in purchasing a Lytro camera that captures the complete light field and would allow adjustments to focus after the image was captured. However the Lytro camera will only export to .jpeg file format.

Do you think it would matter to measure color difference from a .jpeg image? Wouldn't the difference be relative to the color space? He is not trying to get the most accurate color from the images, just measure the difference.

The spiders are shy and move quite quickly so the ability to adjust focus after the fact would be a great advantage.

Thanks

 

[Deleted member 16349]

Not sure where to post this but Color Management seemed the most logical.

I have a biologist friend that is doing research on the change of color of crab spiders measured from digital color images opened within Photoshop. I had suggested to use the RAW file format to eliminate the color conversion that would happen with the jpeg file format. He was interested in purchasing a Lytro camera that captures the complete light field and would allow adjustments to focus after the image was captured. However the Lytro camera will only export to .jpeg file format.

Do you think it would matter to measure color difference from a .jpeg image? Wouldn't the difference be relative to the color space? He is not trying to get the most accurate color from the images, just measure the difference.

The spiders are shy and move quite quickly so the ability to adjust focus after the fact would be a great advantage.

Thanks

One can't obtain accurate colour values with a (rgb) camera. One would need some kind of camera that can take the whole spectrum.

I have read some time back where a special camera with a rotating filter was set used in front of it that could obtain the whole reflected spectrum of an object. Of course it would have to be an object that did not move because of all the different filters required to take all the separate shots (parts of the spectrum).

I believe it was used at RIT in the imaging faculty (not graphics faculty). I think is was used to take images of paintings. Not sure.

 

[gordo]

I have a biologist friend that is doing research on the change of color of crab spiders measured from digital color images opened within Photoshop. I had suggested to use the RAW file format to eliminate the color conversion that would happen with the jpeg file format. He was interested in purchasing a Lytro camera that captures the complete light field and would allow adjustments to focus after the image was captured. However the Lytro camera will only export to .jpeg file format.

Do you think it would matter to measure color difference from a .jpeg image? Wouldn't the difference be relative to the color space? He is not trying to get the most accurate color from the images, just measure the difference.

The spiders are shy and move quite quickly so the ability to adjust focus after the fact would be a great advantage.

Whether the images are jpeg or raw makes no difference.
If he's shooting in an uncontrolled lighting environment there are too many variables that would influence the final color in the image.
So, what he needs is a constant color reference included within the image.
Then he makes any color adjustments needed to make the color reference within the images the same from shot to shot. The Kodak Q-13 gives an idea as to what could work.
Make a small version of the color patches - it can be printed out on an inkjet printer.
The small card with the color patches is the attached to the end of a bar that is attached to the camera. The card should be at a distance from the lens so that it is exposed to the same light that is on the spider. It doesn't have to be where the spider itself is.
In the attached pic the bar holding the color patch target to the camera is that diagonal dark greay fuzzy line.
This is the same basic process that is used on the Mars explorers (and with scanners). The idea being that if the color patches are under the same lighting conditions and made constant in color between different photos then the only difference will be in the color of the subject. Then you can measure and chart the color differences using your preferred method. The Lytro camera would probably be the easiest to set up using this technique because of the shape of the camera itself.

Takes a bit of jerry-rigging but should work.

best, gordo

 

[Deleted member 16349]

This is the same basic process that is used on the Mars explorers (and with scanners). The idea being that if the color patches are under the same lighting conditions and made constant in color between different photos then the only difference will be in the color of the subject.

The pancam for the explorer mission has 16 filters in total, covering most of the visible spectrum. You can not get accurate colour values out of any rgb device no matter what you do. The calibration charts help determine what the light conditions are because the light conditions on Mars are unknown. With the use of the multi filter cameras and the calibration chart, the calibration can be reasonably determined. It could not be with only rgb.

 

[gordo]

The pancam for the explorer mission has 16 filters in total, covering most of the visible spectrum. You can not get accurate colour values out of any rgb device no matter what you do. The calibration charts help determine what the light conditions are because the light conditions on Mars are unknown. With the use of the multi filter cameras and the calibration chart, the calibration can be reasonably determined. It could not be with only rgb.

Yes, that's true for the Mars missions. However, here on Earth, in this particular case, where accuracy is not required but the ability to do a comparison is needed, I think the method I outlined should work.

best, gordo

 

[Deleted member 16349]

Yes, that's true for the Mars missions. However, here on Earth, in this particular case, where accuracy is not required but the ability to do a comparison is needed, I think the method I outlined should work.

best, gordo

Yes, if the level of accuracy does not need to be high. As long as one understands that one is not really comparing colours, then why not.

 

[gordo]

Yes, if the level of accuracy does not need to be high. As long as one understands that one is not really comparing colours, then why not.

Yes, the OP had written: "He is not trying to get the most accurate color from the images, just measure the difference."

In this case he's comparing RGB values (as you say not color) so consistency is more important than accuracy.

best, gordo

 

[Deleted member 16349]

Yes, the OP had written: "He is not trying to get the most accurate color from the images, just measure the difference."

In this case he's comparing RGB values (not color) so consistency is more important than accuracy.

best, gordo

He was talking about colour.

If you scanned a RHEM target, which has stripes that are the same metameric colour, they will most likely come out as different in a scan or a photo done with and rgb device. Would one say that those stripes are different colours? No. They just look different when capturing the image with an rgb device.

My point here is that one can not say anything for sure about colour when using an rgb device.

OK, let him measure the difference in the RGB values. If that satisfies him, I can't argue with that but I hope he is not going to present that as a result based on colour science.

 

[Stephen Marsh]

While reading this topic thread, the following topics came to mind:

Bayer filter - Wikipedia, the free encyclopedia

Foveon X3 sensor - Wikipedia, the free encyclopedia

Lytro - Wikipedia, the free encyclopedia



ColorChecker - Wikipedia, the free encyclopedia



Stephen Marsh

 

[gordo]

My point here is that one can not say anything for sure about colour when using an rgb device.

OK, let him measure the difference in the RGB values. If that satisfies him, I can't argue with that but I hope he is not going to present that as a result based on colour science.

I don't disagree with you. So he'll have to state and qualify the methodology used and the consequent issues in any paper.

best, gordo

 

[hkellogg]

Thanks so much for your comments.

Actually, when I said RGB values, I was thinking I would convert the jpeg images to Lab color space and then use the a, b values for determining our "color difference. I talked to the biologist today and he is really trying to determine how quickly the changes occur so we should be all set.

 

[Deleted member 16349]

Thanks so much for your comments.

Actually, when I said RGB values, I was thinking I would convert the jpeg images to Lab color space and then use the a, b values for determining our "color difference. I talked to the biologist today and he is really trying to determine how quickly the changes occur so we should be all set.

The point of my comments was that one can not convert an rgb value such as with jpeg into Lab values.

Since scientific accuracy is not your goal, I guess you can make some kind of rough relationship to meet you needs.

Good luck.

 

[vedran]

In my experience with image analysis one can pretty accurately determine RGB and Lab values from digital camera images. Not for precise scientific measurements of course. You need to have standardized lightning, camera settings, shoot RAW... There are few ways it can be achieved with pretty good success.

 

[Deleted member 16349]

In my experience with image analysis one can pretty accurately determine RGB and Lab values from digital camera images. Not for precise scientific measurements of course. You need to have standardized lightning, camera settings, shoot RAW... There are few ways it can be achieved with pretty good success.

How do you measure success if you don't compare the results with an accurate colour measurement method?

You are right about it not being a precise scientific measurement and that is my concern. The original poster, I believe, works in a college and I am suspecting that the results might be used in an academic context.

For the average person it is not such an issue since they usually don't follow the science and use experience instead as a guide. Useful most of the time but does not really address all issues.

 

[vedran]

When controlled in the right way I get results equal to spectrophotometer readings (at least within standard tolerances). That might be sufficient for college research if Image Analysis and its applicattion is one of the topics.
Forgot to mention, I compare my results with spectrophotometer readings all the time.

 

[Deleted member 16349]

When controlled in the right way I get results equal to spectrophotometer readings (at least within standard tolerances). That might be sufficient for college research if Image Analysis and its applicattion is one of the topics.
Forgot to mention, I compare my results with spectrophotometer readings all the time.

The problem is in the variety of pigments in Nature. Measuring a relatively few pigments does not validate the method. The results are anecdotal.

It does not concern me how often the method might work in a lab situation or in any anecdotal situation. I look at the science and follow the math. There is no connection between the capture of rgb data and Lab values. To get Lab values, on needs to have captured the spectral curve and apply the xyz functions to get XYZ values which then will lead to Lab values.

By the way, a spectrophotometer does not use standard lighting. You probably know this. It does have to compare what it measures from a target to what it measures from a standard white tile.

If the method was so good, why not use densitometers, which use rgb channels to provide Lab values by some relationship? Well that can not be done even with the limited pigments used in printing but one is expecting that a camera with rgb capture channels is somehow going to be effective in determining Lab values when taking pictures of light from pigments and other lighting phenomena in Nature.

There is concern in the industry that spectrophotometers do not agree so well but the suggestion in the methods approach is that a device that is incapable is somehow going to result in better performance if it uses a special method.

 

[hkellogg]

Maybe I should move away from the concept of measuring color.

The average person might say the spider is changing from white to yellow. This happens when the spider moves from Queen anne's lace to a flower similar to a flower with yellow petals. What is being studied is the rate of change or how fast the change occurs. What we want to know is if the rate is similar, different or the same to change back the "white." For that I would think we could use Photoshop's RGB to Lab conversion with the understanding that this isn't an accurate measure of color but a method to measure the change of one stage to another.

 

[Deleted member 16349]

Maybe I should move away from the concept of measuring color.

The average person might say the spider is changing from white to yellow. This happens when the spider moves from Queen anne's lace to a flower similar to a flower with yellow petals. What is being studied is the rate of change or how fast the change occurs. What we want to know is if the rate is similar, different or the same to change back the "white." For that I would think we could use Photoshop's RGB to Lab conversion with the understanding that this isn't an accurate measure of color but a method to measure the change of one stage to another.

Maybe use a HD video. You could then determine the rate of change. If you clarify how the numbers were obtained and make it clear they are not accurate colour values, I don't see a problem with that.

Also hopefully the spiders are not out of gamut spiders.

 

[meddington]

perhaps a non-contact spectrophotometer could be used?
...if you happened to know of someone nearby who had access to one.