Calculating Area of Ice

Image Analysis
#1

How would I go about accurately calculating the area of ice in this photo https://imgur.com/k6hx1nW

I know that its important to take a good image in the first place but I have a fair few images from past experiments that I dont want to lose.

At the moment I’m having to play around with the contrast/find edges/despeckle/make binary to make the crystals easier to identify and then using morphological segmentation to make an image suitable for thresholding (https://imgur.com/MlAIW2b) . I’ve also tried using Weka segmentation but its eating up my computer’s RAM and doesn’t seem to do a good job.

However I’m using the binary tools a lot to close open crystal boundaries which is leading to me overestimating the mass of solid in the samples.

Are there any better options to what I’m doing?

#2

Please post an uncompressed image, i.e. the original image in either PNG- or TIF-format.

The noise in your image gets worse due to JPG-compression.

Furthermore, it would be interesting to know what accuracy of the area measurement you are expecting.

Regards

Herbie

#3

[1-A] -10 cool - 0091

So I’m trying to keep it within a few % if possible. Im going to try and take more evenly lit colour images but the colour of the two phases Im trying to segment are always similar colour ranges.

#4

Thanks for providing the image data in TIF-format.

While I’m able to achieve a reasonable contour image from the data, a reasonable classification of the ice particles from the background appears out of reach.

I fear that you need a different kind of image acquisition that does not only result in the contours of the ice particles but contrasts their areas.

What kind of optical and illumination process do you presently use during image acquisition?

Regards

Herbie

#5

Its a Leica DM4000 Microscope, with x5, x10, and x25 objectives. Im just using brightfield at the moment. I’m not sure what aspects I should change to increase the contrast of the ice particles?

#6

Thanks for the details!

Obviously you are dealing with near to phase only objects which means that you should use techniques that get intensity contrasts from phase contrasts. I think these techniques are standard — no?

And please re-adjust the illumination!

Regards

Herbie