Help viewing cool flame pressure changes at 3000FPS shadow-graph

I tried the entropy idea cropping the images around the sphere, using a 64x64 pixel thumbnail, and compressing to JPEG.

In the easy test, a statistical change in the image entropy is observed from frame 580.


In the hard test, the sphere apparent position changes (?) so the same analysis fails.


1 Like


here’s difference plot for the easy one

What would think defines the cool flame start?



1 Like

Hello again,
Take at look at these and let me know what you think please. (3.7 MB)

1 Like

@B0RJA @evenhuis

My apologies, I did not explain the problem well enough. The flame I am looking for starts below the droplet about 2 to 3 droplet diameters below it. If you see in this picture there is a rounded wave starting beneath the droplet and it travels towards it well before it ignites

The above image is processed and shows the wave pretty well, below I just drew the behavior of the wave in 4 different times. Trying to illustrate how it goes towards the droplet

So the entropy calculations would be done on a ROI/view box somewhere beneath and slightly to the left of the droplet to catch this cool wave forming.

I’m very amazed still by your results on the droplet. Could you detail the steps you take to get those results? I have never created a plot/graph using ImageJ before.

Thanks so much,

I changed the crop area to difference the images using the previous 50th.

Just plotting the average level of the resulting images shows a consistent increase corresponding to the wave formation.

I include some of the key frames: 300, 325, 350, 375, 400, 425, 450, 475, 500.

I tried to reduce noise to signal ratio, repeating the process using more distant frames (50th, 49th, 48th, 47th, 46) and averaging the results.

Around frame 305 there is an anomaly that could indicate the exact time of the wave formation.

1 Like

Could you list which commands you used to plot average level and in which order you used them in? I can find them in the Menus/Sub-Menus, but I no idea how you are getting these awesome plots!

And regarding your 50th difference, what are you doing exactly? Are you doing difference of every previous 50th frame (The code I showed?) Or just an arbitrarily set previous 50th frame?

Sorry for my confusion, I’d love to know how this process works.


Hello Bob,

These results look really cool! but I am not sure how to interpret them to find the starting frame of the cool flame. I’m also worried this analysis is affected by the blurriness of the image (Due to the half inch of glass the camera shoots through). I detailed the exact phenomenon I’m trying to pin point in a few recent posts. It is very possible I am not understanding the from your results as well.


Hello @evenhuis,

Regarding your first post, when using imageCalculator I tend to prefer using the “Difference” option instead of the “Subtract” option. I do like the idea of adding color however, it seems like some of these LUTs might help me remove some very annoying noise in final product processing.

Also regarding this point, when I do a difference operation the rig still shows up quite a bit. Are you trying to say take some frames well after ignition for the imageCalculator? Also median Z-stack versus average z-stack? :thinking: :thinking: :thinking:

If you could also elaborate on


I am not a ImageJ user but Google gives some similar examples:

I used Imagemagick, a Perl script and Gnuplot to create the graphs:

# loop frames
for($i=10 ; $i<681 ; $i++)
	# crop frame i
	system 'magick convert easy.tif['.$i.'] -crop 256x256+165+480 crop1.tif';

	# crop frame i-10
	system 'magick convert easy.tif['.($i-10).'] -crop 256x256+165+480 crop2.tif';

	# difference frames
	system 'magick composite crop1.tif crop2.tif -compose difference diff.tif';

	# get mean value
	$avg = `magick convert diff.tif -format "%[mean]" info:`;

	print "$i\t$avg\n";

I figured it out! Got better at ImageJ in the process :slight_smile:
This is awesome. Will definitely pursue this avenue further.

Hello again,
The images are the only ones you have sent to work with so I couldn’t back track beyond that point. But assuming a spherical flame front with no external pressure you should be able to use the data (numerically) to back track to a beginning, if any.
Do you think there may have been a mis-calculation in the start times of the clocking sequence? Possibly the ignition occurred prior to image acquisition?

Hello Bob,

Post 20 has all the frames needed to find the wave and then some. Post 24 I did my best to explain the waves I am looking for. The Cool Flame actually starts away from the droplet, so to study the initiation of said flame I look down and to the right. That is what I’m looking to see/study. It starts around slice 300+/- 50 probably. Sorry for the confusion, let me know if I did not clarify well.

FYI due to the free nature of the online transfer service, the files will be deleted in 4 days.


Hello Vasko,
Sorry for the lengthly delay but busy, busy, busy
I don’t know if you found what you were looking for but have a look at this.
Hope it helps,
NewZip to (6.5 MB)

Hey Bob,

B0RJA gave me a very cool technique to find the cool flame with a much higher accuracy so problem solved! However, what you showed me is very interesting! I will try out these “3D Videos”? It seems like an interesting way to see a section over time. Not sure if it will help yet, but you are definitely teaching me more about ImageJ functionality. You know your way around the program for sure!