Vector field quiver plot for sequence

[HamidGadirov]

Hi,

I am wondering how can I use the quiver plot to visualize flow not only for a pair of images but for a sequence where the length of the arrows as well as their color should be relative. For instance, in the plot below all arrows have the same length and color, however, in that simulation the values of the flow in x and y dimensions change, so does the magnitude.

How can this be taken into account and visualized properly using pyimof?

[rfezzani]

Hello @HamidGadirov, and thank you for your interest in pyimof.
I didn't tested it, but I think that you may achieve what you expect by scaling all your displacement fields according to a common scaling factor.

scale_factor = 5  # Arbitrary large enough value
for u, v in OF_SEQ:
    u /= scale_factor
    v /= scale_factor

scale_factor can be the maximum displacement magnitude in the full sequence.

[HamidGadirov]

Hi @rfezzani , thanks for the suggestion, I tried to scale the flow vectors when I am visualizing the results and after I compute the norm, e.g. like this:

u_gt = flow_u_gt
v_gt = flow_v_gt
norm_gt = np.sqrt(u_gt * u_gt + v_gt * v_gt)
pyimof.display.quiver(u_gt / scale_factor, v_gt / scale_factor, c=norm_gt, bg=img, ax=ax, cmap='jet', bg_cmap='gray', step=quiver_steps)

but it didn't lead to any difference in terms of visualization. All vectors still have the same length and red color no matter which scale_factor I have, even a very large one. If I scale before computing the norm, there is no difference either. That changes only if I have different magnitude vectors within each image.

[rfezzani]

After a deeper exploration, it turns out that the solution is simpler then it looks. pyimof.display.quiver accepts any kwarg supported by matplotlib.pyplot.quiver and clim is exactly what you are looking for:

import numpy as np
import pyimof
from matplotlib import pyplot as plt

plt.ion()


u = np.random.rand(460, 640)
v = np.random.rand(460, 640)
c = np.sqrt(u * u + v * v)

alpha = 1.2

ax = pyimof.display.quiver(u, v, c, clim=(-5, 5))
for _ in range(50):
    u *= alpha
    v *= alpha
    c *= alpha
    pyimof.display.quiver(u, v, c, ax=ax, clim=(-5, 5))
    input("Press any key")

[HamidGadirov]

Thanks for your suggestions, looks like using clim improves the visualization of the sequence indeed. I got now the quiver plots with different colors depending on the flow values (I think in this case clim operates with magnitude though, so min. value for clim should be 0). However, after taking a closer look I noticed that the length of the arrows is still the same no matter which flow values are in the corresponding time step:

Do you have any suggestions for getting arrows with different lengths as well?

[rfezzani]

Again, the solution is in matplotlib.pyplot.quiver arguments 😉 : you need to set the scale argument to a value different then None

import numpy as np
import pyimof
from matplotlib import pyplot as plt

plt.ion()


u = np.random.rand(460, 640)
v = np.random.rand(460, 640)
c = np.sqrt(u * u + v * v)

scale = 1

alpha = 1.2

ax = pyimof.display.quiver(u, v, c, clim=(0, 5), scale=scale)
for _ in range(50):
    ax.clear()
    u *= alpha
    v *= alpha
    c *= alpha
    pyimof.display.quiver(u, v, c, ax=ax, clim=(0, 5), scale=scale)
    input("Press any key")

I hope it solves your problem.

[HamidGadirov]

Thanks @rfezzani, this is what I was looking for.