sandplover.mobility.calculate_channelized_response_variance

sandplover.mobility.calculate_channelized_response_variance(arr, threshold=0.2, normalize_input=False, normalize_output=False)

Calculate the Channelized Response Variance (CRV).

This function takes a t-x-y array and calculates its directional CRV [1]. In short, the function does the following:

  1. Normalizes the array at each time slice if desired.

  2. Calculates the CRV magnitude (aka variance along time-axis) and normalizes this array if desired.

  3. Does linear regressions through time for each pixel and returns the slopes.

  4. Calculates the directional CRV using a slope threshold value.

  5. Returns the CRV magnitude, slopes, and directional CRV values

Parameters:

  • arr (numpy.ndarray) – A t-x-y 3-D array to calculate the CRV on.

  • threshold (float, optional) – Threshold for CRV calculation. The default is 0.2.

  • normalize_input (bool, optional) – Whether to normalize the input images pixel values to 0-1. The default is False.

  • normalize_output (bool, optional) – Whether to normalize the output image pixel values to 0-1. The default is False.

Returns:

  • crv_magnitude (numpy.ndarray) – A t-x-y 3-D array with the CRV magnitude.

  • slopes (numpy.ndarray) – A t-x-y 3-D array with the slopes of the linear regressions.

  • directional_crv (numpy.ndarray) – A t-x-y 3-D array with the directional CRV.

Examples

>>> import matplotlib.pyplot as plt
>>> import numpy as np
>>> from sandplover.mobility import calculate_channelized_response_variance
>>> from sandplover.plot import append_colorbar
>>> from sandplover.sample_data.sample_data import savi2020

Load overhead imagery sample data from Savi et al 2020

>>> img, _ = savi2020()

Calculate the CRV on the “Red” band

>>> crv_mag, slopes, crv = calculate_channelized_response_variance(
...     img["red"].data,
...     threshold=0.0,
...     normalize_input=True,
...     normalize_output=True,
... )

Plot the results

>>> fig, ax = plt.subplots(1, 3, figsize=(13, 5))
>>> i0 = ax[0].imshow(crv_mag, vmin=0, vmax=1)
>>> _ = ax[0].set_title("CRV Magnitude")
>>> _ = append_colorbar(i0, ax=ax[0], size=10)
>>> s_ex = np.max([np.abs(slopes.min()), slopes.max()])
>>> i1 = ax[1].imshow(slopes, vmin=-1 * s_ex, vmax=s_ex, cmap="PuOr")
>>> _ = ax[1].set_title("CRV Slopes")
>>> _ = append_colorbar(i1, ax=ax[1], size=10)
>>> i2 = ax[2].imshow(crv, vmin=-1, vmax=1, cmap="seismic")
>>> _ = ax[2].set_title("Directional CRV")
>>> _ = append_colorbar(i2, ax=ax[2], size=10)
>>> _ = fig.suptitle("CRV of Red band from imagery from Savi et al 2020")

(Source code, png, hires.png)

../_images/sandplover-mobility-calculate_channelized_response_variance-1.png