Normalization of True White to Compare Pigment Across Images

I am brand new at imageJ. I’m working on a project where I need to quantify the amount of pigment, regardless of color, in images already taken of flower petals. The pictures were not taken at the same time (i.e. different lighting conditions), but they are all taken using the same platform that the flower sits on. I am trying to set a particular white spot on each image as “true white” so I can compare them. I have found information on how to determine intensity across an area or at a certain point, but I’m just not sure how to “normalize” them so that I can compare them. I uploaded an image as an example. My mentor wants me to use the white word “centimeters” as the place to set as “true white.”

There are commercial color calibration targets. See bhphotovideo for an inexpensive target. I would do some checks on one of those to verify your color response. I would also collect a “background image” from a uniform white target to measure intensity uniformity across the field of view.

You want to verify that you get reproducible results from standards before doing a lot of work on unknowns.

Store the targets in the dark in the bags they come in. Colors fade and change with continued exposure to light. Some are sensitive to plasticizers (so use the bags the vendor supplies). My internal clients at Kodak were color experts and stressed the importance of careful handling and exposure…

Best Regards,
John Minter
Retired from Kodak Analytical Sciences

Thank you so much for your reply.

Do you need to use the color calibration targets as you take original images? I have just started in a lab and I have received a folder of several hundred images like the one I uploaded to analyze. We no longer have the flower samples, just the images.

I have a macro that corrects white balance using a reference ROI (supposed to be neutral) in the input image.

It’s not perfect but works in many cases. Here’s an example on the white balance test image distributed with the simplecv package.

image

Jerome,

This is exactly what I looking for! Thank you very much

1 Like