def lbp():
# Import dataset
X, Y, lb = Data.numbers()
X_new = []
for img in X:
# Convert array in image
image = img.reshape([35, 35])
image = image.astype(np.uint8)
# Number of points to be considered as neighbourers
radius = 3
no_points = 8 * radius
# Uniform LBP is used
lbp = local_binary_pattern(image,
no_points,
radius,
method='uniform')
lbp = np.array(lbp).flatten()
#print(lbp)
X_new.append(lbp)
return np.array(X_new), Y, lb
def glcm():
# Import dataset
X, Y, lb = Data.numbers()
X_new = []
for img in X:
# Convert array in image
image = img.reshape([35, 35])
image = image.astype(np.uint8)
g = greycomatrix(image, [0, 1], [0, np.pi / 2], levels=256)
contrast = greycoprops(g, 'contrast').flatten()
energy = greycoprops(g, 'energy').flatten()
homogeneity = greycoprops(g, 'homogeneity').flatten()
correlation = greycoprops(g, 'correlation').flatten()
dissimilarity = greycoprops(g, 'dissimilarity').flatten()
ASM = greycoprops(g, 'ASM').flatten()
features = np.concatenate((contrast, energy, homogeneity,
correlation, dissimilarity, ASM))
X_new.append(features)
return X_new, Y, lb
def huMoments():
# Import dataset
X, Y, lb = Data.numbers()
X_new = []
for img in X:
# Convert array in image
image = img.reshape([35, 35])
image = image.astype(np.uint8)
# Calculate Moments
moments = cv2.moments(image)
# Calculate Hu Moments
huMoments = cv2.HuMoments(moments)
# Log scale hu moments
hm = [
-1 * copysign(1.0, hu) * log10(abs(hu)) if hu != 0 else 0
for hu in huMoments
]
X_new.append(hm)
return np.array(X_new), Y, lb