def getmultiLBP(mat,step=50,block=100):
N=mat.shape[0]
idx=np.triu_indices(N)
lbp = []
for i,j in itertools.product(idx[0],idx[1]):
lbp.append(feature.multiblock_lbp(mat,i,j,block,block))
colNames = [ 'lbp_rad_%d' %(r) for r in range(len(lbp))]
lbpSeries = pd.Series(data = preprocessing.scale(lbp), index=colNames)
return lbpSeries
def test_single_mblbp(self):
# Create dummy matrix where first and fifth rectangles have greater
# value than the central one. Therefore, the following bits
# should be 1.
test_img = np.zeros((9, 9), dtype='uint8')
test_img[3:6, 3:6] = 1
test_img[:3, :3] = 255
test_img[6:, 6:] = 255
# MB-LBP is filled in reverse order. So the first and fifth bits from
# the end should be filled.
correct_answer = 0b10001000
int_img = integral_image(test_img)
lbp_code = multiblock_lbp(int_img, 0, 0, 3, 3)
np.testing.assert_equal(lbp_code, correct_answer)
def test_single_mblbp(self):
# Create dummy matrix where first and fifth rectangles have greater
# value than the central one. Therefore, the following bits
# should be 1.
test_img = np.zeros((9, 9), dtype='uint8')
test_img[3:6, 3:6] = 1
test_img[:3, :3] = 255
test_img[6:, 6:] = 255
# MB-LBP is filled in reverse order. So the first and fifth bits from
# the end should be filled.
correct_answer = 0b10001000
int_img = integral_image(test_img)
lbp_code = multiblock_lbp(int_img, 0, 0, 3, 3)
np.testing.assert_equal(lbp_code, correct_answer)
def histogram_feature(real_img, img, p_blocks, n_features, h_size, slider):
xsize, ysize = img.shape
histogram = np.zeros(h_size, dtype=float)
y = 0
x = 0
while ((y + 3 * slider) <= ysize):
while ((x + 3 * slider) <= xsize):
lbp_code = multiblock_lbp(img, y, x, p_blocks, p_blocks)
histogram[lbp_code] += 1
x += slider
#region_print(real_img, lbp_code, y+20, x+15, p_blocks, p_blocks)
y += slider
x = 0
histogram = np.sort(histogram, axis=None)
features = np.zeros(n_features, dtype=float)
histogram[:h_size - n_features]
features[0:n_features - 1] = histogram[h_size - (n_features - 1):]
features[n_features - 1] = np.sum(histogram[:h_size - (n_features - 1)])
max = np.amax(features)
return features / max
def calculate_lbp(img_path):
# load img in grayscale
img = load(img_path, as_gray=True)
return multiblock_lbp(img, 0, 0, WIDTH, HEIGHT)
from __future__ import print_function from skimage.feature import multiblock_lbp, draw_multiblock_lbp import numpy as np from numpy.testing import assert_equal from skimage.transform import integral_image from matplotlib import pyplot as plt # Create test matrix where first and fifth rectangles starting # from top left clockwise have greater value than the central one. test_img = np.zeros((9, 9), dtype='uint8') test_img[3:6, 3:6] = 1 test_img[:3, :3] = 50 test_img[6:, 6:] = 50 print(test_img) # First and fifth bits should be filled. This correct value will # be compared to the computed one. correct_answer = 0b10001000 int_img = integral_image(test_img) print(int_img) lbp_code = multiblock_lbp(int_img, 0, 0, 3, 3) print(lbp_code) img = draw_multiblock_lbp(test_img, 0, 0, 90, 90, lbp_code=lbp_code, alpha=0.5) plt.imshow(img, interpolation='nearest') plt.show()
def getFeatures_old(I, Isegmented):
Ilab = rgb2lab(I)
Ilab[:, :, 0] *= 2.55
Ilab[:, :, 1] += 127
Ilab[:, :, 2] += 128
I05 = rescale(I, 0.5)
I025 = rescale(I, 0.25)
I0125 = rescale(I, 0.125)
Ilab05 = rescale(Ilab, 0.5)
Ilab025 = rescale(Ilab, 0.25)
Ilab0125 = rescale(Ilab, 0.125)
Isegmented05 = rescale(Isegmented, 0.5)
Isegmented025 = rescale(Isegmented, 0.25)
Isegmented0125 = rescale(Isegmented, 0.125)
Xstack = []
Xnstack = []
options = {'b': [
{'name': 'basicgeo', 'options': {'show': False}}, # basic geometric features
{'name': 'hugeo', 'options': {'show': False}}, # Hu moments
{'name': 'flusser', 'options': {'show': False}}, # Flusser moments
{'name': 'fourierdes', 'options': {'show': False, 'Nfourierdes': 12}}, # Fourier descriptors
]}
Xtmp, Xntmp = Bfx_geo(Isegmented, options)
Xstack.extend(Xtmp[0])
Xnstack.extend(Xntmp)
options = {'dharalick': [3, 6, 12]} # # pixels distance for coocurrence
# options = {'dharalick': 3} # 3 pixels distance for coocurrence
J = rgb2gray(I.astype(float))
Xtmp, Xntmp = Bfx_haralick(J, Isegmented, options) # Haralick features
Xntmp = [name + '_gray' for name in Xntmp]
Xstack.extend(Xtmp[0])
Xnstack.extend(Xntmp)
Xtmp, Xntmp = Bfx_haralick(Ilab[:, :, 0], Isegmented, options) # Haralick features
Xntmp = [name + '_L*' for name in Xntmp]
Xstack.extend(Xtmp[0])
Xnstack.extend(Xntmp)
Xtmp, Xntmp = Bfx_haralick(Ilab[:, :, 1], Isegmented, options) # Haralick features
Xntmp = [name + '_A*' for name in Xntmp]
Xstack.extend(Xtmp[0])
Xnstack.extend(Xntmp)
Xtmp, Xntmp = Bfx_haralick(Ilab[:, :, 2], Isegmented, options) # Haralick features
Xntmp = [name + '_B*' for name in Xntmp]
Xstack.extend(Xtmp[0])
Xnstack.extend(Xntmp)
a, _ = histogram(np.round(rgb2gray(I.astype(float))))
a = a / a.sum()
Xtmp = [entr(a).sum(axis=0)]
Xntmp = ['Entropy']
Xstack.extend(Xtmp)
Xnstack.extend(Xntmp)
count = float(I[:, :, 0].size)
mean_red = np.sum(I[:, :, 0])
mean_green = np.sum(I[:, :, 1])
mean_blue = np.sum(I[:, :, 2])
Xtmp = [mean_red / count, mean_green / count, mean_blue / count]
Xntmp = ['mean_red', 'mean_green', 'mean_blue']
Xstack.extend(Xtmp)
Xnstack.extend(Xntmp)
Xstack.extend(Xtmp)
Xnstack.extend(Xntmp)
mean_l = np.sum(Ilab[:, :, 0])
mean_a = np.sum(Ilab[:, :, 1])
mean_b = np.sum(Ilab[:, :, 2])
Xtmp = [mean_l / count, mean_a / count, mean_b / count]
Xntmp = ['mean_l', 'mean_a', 'mean_b']
Xstack.extend(Xtmp)
Xnstack.extend(Xntmp)
options = {
'weight': 0,
'vdiv': 3,
'hdiv': 3,
'samples': 8,
'mappingtype': 'nri_uniform'
}
Xtmp, Xntmp = Bfx_lbp(I[:, :, 0], Isegmented, options)
Xntmp = [name + '_red_normal' for name in Xntmp]
Xstack.extend(Xtmp[0])
Xnstack.extend(Xntmp)
Xtmp, Xntmp = Bfx_lbp(I[:, :, 1], Isegmented, options)
Xntmp = [name + '_green_normal' for name in Xntmp]
Xstack.extend(Xtmp[0])
Xnstack.extend(Xntmp)
Xtmp, Xntmp = Bfx_lbp(I[:, :, 2], Isegmented, options)
Xntmp = [name + '_blue_normal' for name in Xntmp]
Xstack.extend(Xtmp[0])
Xnstack.extend(Xntmp)
Xtmp, Xntmp = Bfx_lbp(rgb2gray(I), Isegmented, options)
Xntmp = [name + '_gray_normal' for name in Xntmp]
Xstack.extend(Xtmp[0])
Xnstack.extend(Xntmp)
Xtmp, Xntmp = Bfx_lbp(Ilab[:, :, 0], Isegmented, options)
Xntmp = [name + '_L*_normal' for name in Xntmp]
Xstack.extend(Xtmp[0])
Xnstack.extend(Xntmp)
Xtmp, Xntmp = Bfx_lbp(Ilab[:, :, 1], Isegmented, options)
Xntmp = [name + '_A*_normal' for name in Xntmp]
Xstack.extend(Xtmp[0])
Xnstack.extend(Xntmp)
Xtmp, Xntmp = Bfx_lbp(Ilab[:, :, 2], Isegmented, options)
Xntmp = [name + '_B*_normal' for name in Xntmp]
Xstack.extend(Xtmp[0])
Xnstack.extend(Xntmp)
# Multiblock_LBP - RGB - normal
ILabel = label(Isegmented)
for region in regionprops(ILabel):
minr, minc, maxr, maxc = region.bbox
Xtmp = multiblock_lbp(I[:, :, 0], minr, minc, int((maxc - minc) / 3), int((maxr - minr) / 3))
Xntmp = 'multiblock_lbp_red_normal'
Xstack.extend([Xtmp])
Xnstack.extend([Xntmp])
Xtmp = multiblock_lbp(I[:, :, 1], minr, minc, int((maxc - minc) / 3), int((maxr - minr) / 3))
Xntmp = 'multiblock_lbp_green_normal'
Xstack.extend([Xtmp])
Xnstack.extend([Xntmp])
Xtmp = multiblock_lbp(I[:, :, 2], minr, minc, int((maxc - minc) / 3), int((maxr - minr) / 3))
Xntmp = 'multiblock_lbp_blue_normal'
Xstack.extend([Xtmp])
Xnstack.extend([Xntmp])
# Multiblock_LBP - Gray - normal
Xtmp = multiblock_lbp(rgb2gray(I), minr, minc, int((maxc - minc) / 3), int((maxr - minr) / 3))
Xntmp = 'multiblock_lbp_gray_normal'
Xstack.extend([Xtmp])
Xnstack.extend([Xntmp])
# Multiblock_LBP - LBP - normal
Xtmp = multiblock_lbp(Ilab[:, :, 0], minr, minc, int((maxc - minc) / 3), int((maxr - minr) / 3))
Xntmp = 'multiblock_lbp_L*_normal'
Xstack.extend([Xtmp])
Xnstack.extend([Xntmp])
Xtmp = multiblock_lbp(Ilab[:, :, 1], minr, minc, int((maxc - minc) / 3), int((maxr - minr) / 3))
Xntmp = 'multiblock_lbp_A*_normal'
Xstack.extend([Xtmp])
Xnstack.extend([Xntmp])
Xtmp = multiblock_lbp(Ilab[:, :, 2], minr, minc, int((maxc - minc) / 3), int((maxr - minr) / 3))
Xntmp = 'multiblock_lbp_B*_normal'
Xstack.extend([Xtmp])
Xnstack.extend([Xntmp])
# Multiblock_LBP - RGB - 0.5
ILabel = label(Isegmented05)
for region in regionprops(ILabel):
minr, minc, maxr, maxc = region.bbox
Xtmp = multiblock_lbp(I05[:, :, 0], minr, minc, int((maxc - minc) / 3), int((maxr - minr) / 3))
Xntmp = 'multiblock_lbp_red_0.5'
Xstack.extend([Xtmp])
Xnstack.extend([Xntmp])
Xtmp = multiblock_lbp(I05[:, :, 1], minr, minc, int((maxc - minc) / 3), int((maxr - minr) / 3))
Xntmp = 'multiblock_lbp_green_0.5'
Xstack.extend([Xtmp])
Xnstack.extend([Xntmp])
Xtmp = multiblock_lbp(I05[:, :, 2], minr, minc, int((maxc - minc) / 3), int((maxr - minr) / 3))
Xntmp = 'multiblock_lbp_blue_0.5'
Xstack.extend([Xtmp])
Xnstack.extend([Xntmp])
# Multiblock_LBP - Gray - normal
Xtmp = multiblock_lbp(rgb2gray(I05), minr, minc, int((maxc - minc) / 3), int((maxr - minr) / 3))
Xntmp = 'multiblock_lbp_gray_0.5'
Xstack.extend([Xtmp])
Xnstack.extend([Xntmp])
# Multiblock_LBP - LBP - normal
Xtmp = multiblock_lbp(Ilab05[:, :, 0], minr, minc, int((maxc - minc) / 3), int((maxr - minr) / 3))
Xntmp = 'multiblock_lbp_L*_0.5'
Xstack.extend([Xtmp])
Xnstack.extend([Xntmp])
Xtmp = multiblock_lbp(Ilab05[:, :, 1], minr, minc, int((maxc - minc) / 3), int((maxr - minr) / 3))
Xntmp = 'multiblock_lbp_A*_0.5'
Xstack.extend([Xtmp])
Xnstack.extend([Xntmp])
Xtmp = multiblock_lbp(Ilab05[:, :, 2], minr, minc, int((maxc - minc) / 3), int((maxr - minr) / 3))
Xntmp = 'multiblock_lbp_B*_0.5'
Xstack.extend([Xtmp])
Xnstack.extend([Xntmp])
# Multiblock_LBP - RGB - 0.25
ILabel = label(Isegmented025)
for region in regionprops(ILabel):
minr, minc, maxr, maxc = region.bbox
Xtmp = multiblock_lbp(I025[:, :, 0], minr, minc, int((maxc - minc) / 3), int((maxr - minr) / 3))
Xntmp = 'multiblock_lbp_red_0.25'
Xstack.extend([Xtmp])
Xnstack.extend([Xntmp])
Xtmp = multiblock_lbp(I025[:, :, 1], minr, minc, int((maxc - minc) / 3), int((maxr - minr) / 3))
Xntmp = 'multiblock_lbp_green_0.25'
Xstack.extend([Xtmp])
Xnstack.extend([Xntmp])
Xtmp = multiblock_lbp(I025[:, :, 2], minr, minc, int((maxc - minc) / 3), int((maxr - minr) / 3))
Xntmp = 'multiblock_lbp_blue_0.25'
Xstack.extend([Xtmp])
Xnstack.extend([Xntmp])
# Multiblock_LBP - Gray - normal
Xtmp = multiblock_lbp(rgb2gray(I025), minr, minc, int((maxc - minc) / 3), int((maxr - minr) / 3))
Xntmp = 'multiblock_lbp_gray_0.25'
Xstack.extend([Xtmp])
Xnstack.extend([Xntmp])
# Multiblock_LBP - LBP - normal
Xtmp = multiblock_lbp(Ilab025[:, :, 0], minr, minc, int((maxc - minc) / 3), int((maxr - minr) / 3))
Xntmp = 'multiblock_lbp_L*_0.25'
Xstack.extend([Xtmp])
Xnstack.extend([Xntmp])
Xtmp = multiblock_lbp(Ilab025[:, :, 1], minr, minc, int((maxc - minc) / 3), int((maxr - minr) / 3))
Xntmp = 'multiblock_lbp_A*_0.25'
Xstack.extend([Xtmp])
Xnstack.extend([Xntmp])
Xtmp = multiblock_lbp(Ilab025[:, :, 2], minr, minc, int((maxc - minc) / 3), int((maxr - minr) / 3))
Xntmp = 'multiblock_lbp_B*_0.25'
Xstack.extend([Xtmp])
Xnstack.extend([Xntmp])
# Multiblock_LBP - RGB - 0.125
ILabel = label(Isegmented0125)
for region in regionprops(ILabel):
minr, minc, maxr, maxc = region.bbox
Xtmp = multiblock_lbp(I0125[:, :, 0], minr, minc, int((maxc - minc) / 3), int((maxr - minr) / 3))
Xntmp = 'multiblock_lbp_red_0.125'
Xstack.extend([Xtmp])
Xnstack.extend([Xntmp])
Xtmp = multiblock_lbp(I0125[:, :, 1], minr, minc, int((maxc - minc) / 3), int((maxr - minr) / 3))
Xntmp = 'multiblock_lbp_green_0.125'
Xstack.extend([Xtmp])
Xnstack.extend([Xntmp])
Xtmp = multiblock_lbp(I0125[:, :, 2], minr, minc, int((maxc - minc) / 3), int((maxr - minr) / 3))
Xntmp = 'multiblock_lbp_blue_0.125'
Xstack.extend([Xtmp])
Xnstack.extend([Xntmp])
# Multiblock_LBP - Gray - normal
Xtmp = multiblock_lbp(rgb2gray(I0125), minr, minc, int((maxc - minc) / 3), int((maxr - minr) / 3))
Xntmp = 'multiblock_lbp_gray_0.125'
Xstack.extend([Xtmp])
Xnstack.extend([Xntmp])
# Multiblock_LBP - LBP - normal
Xtmp = multiblock_lbp(Ilab0125[:, :, 0], minr, minc, int((maxc - minc) / 3), int((maxr - minr) / 3))
Xntmp = 'multiblock_lbp_L*_0.125'
Xstack.extend([Xtmp])
Xnstack.extend([Xntmp])
Xtmp = multiblock_lbp(Ilab0125[:, :, 1], minr, minc, int((maxc - minc) / 3), int((maxr - minr) / 3))
Xntmp = 'multiblock_lbp_A*_0.125'
Xstack.extend([Xtmp])
Xnstack.extend([Xntmp])
Xtmp = multiblock_lbp(Ilab0125[:, :, 2], minr, minc, int((maxc - minc) / 3), int((maxr - minr) / 3))
Xntmp = 'multiblock_lbp_B*_0.125'
Xstack.extend([Xtmp])
Xnstack.extend([Xntmp])
return Xstack, Xnstack
from skimage.transform import integral_image # Create test matrix where first and fifth rectangles starting # from top left clockwise have greater value than the central one. test_img = np.zeros((9, 9), dtype='uint8') test_img[3:6, 3:6] = 1 test_img[:3, :3] = 50 test_img[6:, 6:] = 50 # First and fifth bits should be filled. This correct value will # be compared to the computed one. correct_answer = 0b10001000 int_img = integral_image(test_img) lbp_code = multiblock_lbp(int_img, 0, 0, 3, 3) assert_equal(correct_answer, lbp_code) ###################################################################### # Now let's apply the operator to a real image and see how the visualization # works. from skimage import data from matplotlib import pyplot as plt from skimage.feature import draw_multiblock_lbp test_img = data.coins() int_img = integral_image(test_img)
def compareScikit():
test_img = np.round(np.random.rand(36, 36) * 256)
int_img = integral_image(test_img)
matriks = genAverageMat(test_img, 2, 3)
mat_lbp = lbpCompare(matriks)
lbp_2 = multiblock_lbp(int_img, 0, 9, 3, 2)
from skimage import data
from matplotlib import pyplot as plt
from skimage.transform import integral_image
from skimage.feature import multiblock_lbp
from skimage.feature import draw_multiblock_lbp
test_img = data.coins()
plt.imshow(test_img, interpolation='nearest')
plt.show()
int_img = integral_image(test_img)
lbp_code = multiblock_lbp(int_img, 0, 0, 190, 90)
img = draw_multiblock_lbp(test_img,
0,
0,
190,
90,
lbp_code=lbp_code,
alpha=0.5)
plt.imshow(img, interpolation='nearest')
plt.show()