1

I am trying to calculate the texture features of my 1 image using GLCM. Now if I use these properties properties = ['contrast', 'energy','ASM', 'homogeneity', 'correlation', 'dissimilarity'] and calculate GLCM in 4 direction angles like [0, np.pi / 4, np.pi / 2, 3 * np.pi / 4] then I get 4 *6=24 feature values. Now my professor is saying 24 features are not enough for the feature selection so calculate GLCM in more directions to get more feature values like previously my directions were [0, 45,90,135] now he is asking me to calculate in [120, 180, 270], etc just to get more features. But everywhere on google I have seen people are calculating GLCM in the standard four directions that is [0, np.pi / 4, np.pi / 2, 3 * np.pi / 4] so is this correct if I calculate GLCM in the other directions which he mentioned?

kiran
  • 11
  • 2
  • Why not? You only take more features into consideration that way. I. the worst case scenario they aren't much more useful than the 4 directional features and you could even compare them and argue against it if it applies. No reason to not try this out. – T A Nov 10 '20 at 11:53
  • @ T A Actually, everywhere on google, I found researches are only using those standard 4 angles that is why I was confused. Between I am trying to find different properties like entropy, sum average, sum variance then the standard 6 properties to get more features instead of adding more angles but I am getting the error do you know why is that? Is that because greycoprops does not calculate these properties? print(skimage.feature.greycoprops(g, 'sum variance ')]) print(skimage.feature.greycoprops(g, 'entropy')) print(skimage.feature.greycoprops(g, 'Cluster Shade')) – kiran Nov 10 '20 at 13:55
  • Which error? Please add you code in your question instead of the comments, this way you can format it. – T A Nov 11 '20 at 13:19

0 Answers0