Identify a weak and fast-moving pattern from raw images?

Questions
1

I am currently doing post-processing of a weak and fast-moving pattern (actually waves, if you are familiar with Schlieren Imaging…)

Now the problem is that, other than by directly showing the raw images and measure by hand, I have no good ways to visualize it neatly or extract any data from it (the main point of the post-processing is to find out the frequency! Or even more ambitiously, the velocity of the wave moving…)

In a nutshell, what I have in mind about the problem is that:

The patterns are weak, hence any operation of noise removal would hurt it

The background is uneven and noisy

The patterns move too quickly: within four or five frames the patterns would pass by the whole range of image

Here are an example of images: I have selected 3 that are representative.

stat_1.png
stat_1.png1024×1024 44.6 KB

stat_2.png
stat_2.png1024×1024 47.2 KB

stat_3.png
stat_3.png1024×1024 51.2 KB

For better understanding, I have made an animation with gif out of the results:

Anim.gif

2

Thanks everyone for the possible help :stuck_out_tongue:

3

Maybe take a Fourier transform at every pixel? That is, treat each pixel value as a function of time and take the FT of a the sequence of image gray scale values at that pixel. Then average over all pixels. This can perhaps give an estimate of frequency and speed.

4

Hi Sihe Chen,

this might be just a first step; the basic idea is to emphasize vertical structures by using GradientOrientationFilter. So at least one has the impression that “something is passing by”. In the following I am using the images from your animated gif, because this contains 5 images instead of only 3.

imgs0 = Import[
   "https://community.wolfram.com//c/portal/getImageAttachment?\
filename=Anim.gif&userId=1604256"];
imgs1 = ImageAdjust@*
    RemoveAlphaChannel@*(ColorConvert[#, "Grayscale"] &) /@ imgs0;
imgs2 = ImageAdjust@
     ImageFilter[MedianDeviation[Flatten[#]] &, #, 6] & /@ imgs1;
imgs3 = ImageApply[If[# > 1, 1, 0] &, 
     GradientOrientationFilter[Binarize[#, .3], 2]] & /@ imgs2;
imgs4 = ImageAdjust[Dilation[Erosion[Blur[#, 5], 2], 2]] & /@ imgs3

This gives:

ttt1.gif

I admit that this is not really satisfying. Regards – Henrik
[1]: https://community.wolfram.com//c/portal/getImageAttachment?filename=ttt1.gif&userId=32203

5

Thanks Daniel, but due to the noise and fast movement Fourier transformation is overwhelmed by the noise in the power spectrum…

6

Thanks Henrik! This seems to make a huge advance. I will try and see whether any further processing could be made based on the current result.