Added documentation for Fourier-based methods

Author: zyjux

src/sharpness/fourier.py

@@ -1,7 +1,10 @@
+"""Module containing low-level functions based on Fourier analysis"""
+
 import numpy as np
 
 
 def compute_power_spectrum(image, hanning=True):
+    """Utility method used for the other metrics in this"""
     N = image.shape[0]
     if hanning:
         # Set up 2D Hanning window to deal with edge effects
@@ -16,16 +19,18 @@ def compute_power_spectrum(image, hanning=True):
 
 
 def fourier_rmse(image1, image2, hanning=True):
+    """Bivariate -- Unweighted RMSE between power spectra of the input images"""
     # Compute power spectra of both images
     power_spectrum1 = compute_power_spectrum(image1, hanning=hanning)
     power_spectrum2 = compute_power_spectrum(image2, hanning=hanning)
 
     # Compute the mean squared error between power spectra
-    mse = np.mean((power_spectrum1-power_spectrum2)**2)
+    mse = np.mean((power_spectrum1 - power_spectrum2) ** 2)
     return np.sqrt(mse)
 
 
 def fourier_total_variation(image, hanning=True):
+    """Univariate -- Total variation within the power spectra of a given image"""
     N = image.shape[0]
     if hanning:
         # Set up 2D Hanning window to deal with edge effects
@@ -36,12 +41,3 @@ def fourier_total_variation(image, hanning=True):
     f_transform = np.fft.fft2(image)
     tv = np.sum(np.abs(f_transform))
     return tv
-
-
-if __name__ == '__main__':
-    from skimage.data import camera
-    image1 = camera()
-    image2 = camera()
-
-    fourier_rmse = fourier_rmse(image1, image2)
-    print("Fourier power spectrum RMSE:", fourier_rmse)