update

Beethoven.py

@@ -1,17 +1,18 @@
 # coding: utf-8
 from numpy import *
 from processing import *
-import matplotlib.pyplot as plt 
+import matplotlib.pyplot as plt
 
 # read the .mat file
-I,mask,S = read_data_file("Data/Beethoven.mat")
+I, mask, S = read_data_file("Data/Beethoven.mat")
 
 for i in range(I.shape[2]):
-    plt.subplot(1,3,i+1)
-    plt.imshow(I[:,:,i],cmap = plt.cm.gray)
+    plt.subplot(1, 3, i+1)
+    plt.imshow(I[:, :, i], cmap = plt.cm.gray)
     plt.axis('off')
 
-I_all = vstack((I[:,:,0].flatten(),I[:,:,1].flatten(),I[:,:,2].flatten()))
+I_all = vstack(
+    (I[:,:,0].flatten(), I[:,:,1].flatten(), I[:,:,2].flatten()))
 
 M = linalg.inv(S).dot(I_all)
 M_1 = M[0]

README.md

@@ -9,7 +9,7 @@ camera angle but with varying light angles. According to Lambert’s law, for a
 
 <img src='/src/1.png' width='400'>
 
-Where I is the vector of k observed intensities, S is the known 3 x k matrix of normalized light directions, and M is the surface normal that we need. The
+Where I is the vector of k observed intensities, S is the known k x 3 matrix of normalized light directions, and M is the surface normal that we need. The
 algorithms is then for each (valid) pixel [u, v] in image domain, solve m(u, v)
 via Moore-Penrose pseudoinverse or equation selection. Then we get the albedo and normal with the following equation:
 

processing/ps_utils.py

@@ -12,6 +12,11 @@
 from scipy.io import loadmat
 from matplotlib import pyplot as plt
 
+try:
+    xrange
+except NameError:
+    xrange = range
+
 def cdx(f):
     """
     central differences for f-
@@ -196,8 +201,8 @@ def unbiased_integrate(n1, n2, n3, mask, order=2):
         pbar = p.copy()
         qbar = q.copy()
     elif order == 2:
-        pbar = 0.5*(p + p[range(1,m) + [m-1], :])
-        qbar = 0.5*(q + q[:, range(1,n) + [n-1]])
+        pbar = 0.5*(p + p[list(range(1,m)) + [m-1], :])
+        qbar = 0.5*(q + q[:, list(range(1,n)) + [n-1]])
 
     # System
     I = []