def showSobelLaplacian(segmentation_type, input_img):
if (segmentation_type != "none"):
img = cv2.imread(input_img, 0) # uint8 image in grayscale
else:
img = cv2.imread(input_img)
results = list()
img = cv2.normalize(img, None, 0, 255, cv2.NORM_MINMAX)
results.append(img)
if (segmentation_type == "otsu"):
segmented_img = processedSegmentation.imgSegmentation(img)
elif (segmentation_type == "gauss"):
segmented_img = processedSegmentation.adaptiveSegmentationGaussian(img)
elif (segmentation_type == "mean"):
segmented_img = processedSegmentation.adaptiveSegmentationMean(img)
if (segmentation_type != "none"):
cv2.imwrite('./processedImg/segmented_img.jpg', segmented_img)
img = cv2.imread('./processedImg/segmented_img.jpg')
img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
laplacian = cv2.Laplacian(img,
cv2.CV_64F) # processed with Laplacian operator
cv2.imwrite("./processedImg/laplacian_img.jpg", laplacian)
laplacian_img = cv2.imread("./processedImg/laplacian_img.jpg", 0)
results.append(laplacian_img)
sobelx = cv2.Sobel(img, cv2.CV_64F, 1, 0,
ksize=5) # processed with Sobel on x-axis
cv2.imwrite("./processedImg/sobelx_img.jpg", sobelx)
sobelx_img = cv2.imread("./processedImg/sobelx_img.jpg", 0)
results.append(sobelx_img)
sobely = cv2.Sobel(img, cv2.CV_64F, 0, 1,
ksize=5) # processed with Sobel on y-axis
cv2.imwrite("./processedImg/sobely_img.jpg", sobely)
sobely_img = cv2.imread("./processedImg/sobely_img.jpg", 0)
results.append(sobely_img)
titles = [
'Original normalized image', 'Applied Laplacian operator',
'Applied Sobel operator of x axis', 'Applied Sobel operator of y axis'
]
fig = plt.figure(figsize=(11, 11))
i = 0
for a in results:
ax = fig.add_subplot(2, 2, i + 1)
ax.imshow(a, cmap='gray')
ax.set_title(titles[i], fontsize=10)
ax.set_xticks([])
ax.set_yticks([])
i = i + 1
fig.tight_layout()
plt.show()
return
def showLBP(segmentation_type, input_img):
if (segmentation_type != "none"):
img = cv2.imread(input_img, 0) # uint8 image in grayscale
else:
img = cv2.imread(input_img)
#img = cv2.resize(img,(400,400)) # resize of image
img = cv2.normalize(img, None, 0, 255, cv2.NORM_MINMAX) # normalize image
#segmented_img = adaptiveSegmentationMean(img)
if (segmentation_type == "otsu"):
segmented_img = processedSegmentation.imgSegmentation(img)
elif (segmentation_type == "gauss"):
segmented_img = processedSegmentation.adaptiveSegmentationGaussian(img)
elif (segmentation_type == "mean"):
segmented_img = processedSegmentation.adaptiveSegmentationMean(img)
#cv2.imshow('Segmented image', segmented_img)
if (segmentation_type != "none"):
cv2.imwrite('./processedImg/segmented_img.jpg', segmented_img)
img = cv2.imread('./processedImg/segmented_img.jpg')
height, width, channel = img.shape
img_gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
lbp_image = np.zeros((height, width, 3), np.uint8)
# processing LBP algorithm
lbp_values = []
for i in range(0, height):
for j in range(0, width):
lbp_image[i, j] = processLBP(img_gray, i, j, lbp_values)
cv2.imshow("lbp image", lbp_image)
hist_lbp = cv2.calcHist([lbp_image], [0], None, [256], [0, 256])
figure = plt.figure(figsize=(30, 30))
image_plot = figure.add_subplot(1, 2, 1)
image_plot.imshow(lbp_image)
image_plot.set_xticks([])
image_plot.set_yticks([])
image_plot.set_title("LBP image", fontsize=10)
current_plot = figure.add_subplot(1, 2, 2)
current_plot.plot(hist_lbp, color=(0, 0, 0.2))
#current_plot.set_xlim([0,260])
current_plot.set_xlim([0, 256])
current_plot.set_ylim([0, 6000])
current_plot.set_title("LBP histogram", fontsize=10)
current_plot.set_xlabel("Intensity")
current_plot.set_ylabel("Count of pixels")
ytick_list = [int(i) for i in current_plot.get_yticks()]
current_plot.set_yticklabels(ytick_list, rotation=90)
plt.show()
cv2.waitKey(0)
cv2.destroyAllWindows()
return
def showWavelet(segmentation_type, input_img, wavelet_type):
if (segmentation_type != "none"):
img = cv2.imread(input_img, 0) # uint8 image in grayscale
else:
img = cv2.imread(input_img)
img = cv2.normalize(img, None, 0, 255, cv2.NORM_MINMAX) # normalize image
# choose type of segmentation
if (segmentation_type == "otsu"):
segmented_img = processedSegmentation.imgSegmentation(img)
elif (segmentation_type == "gauss"):
segmented_img = processedSegmentation.adaptiveSegmentationGaussian(img)
elif (segmentation_type == "mean"):
segmented_img = processedSegmentation.adaptiveSegmentationMean(img)
if (segmentation_type != "none"):
cv2.imwrite('./processedImg/segmented_img.jpg', segmented_img)
img = cv2.imread('./processedImg/segmented_img.jpg')
img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
# Convert to float32 for more resolution for use with pywt
img = np.float32(img)
img /= 255
titles = [
'Approximation (LL)', ' Horizontal detail (LH)',
'Vertical detail (HL)', 'Diagonal detail (HH)'
]
coeffs2 = pywt.dwt2(img, wavelet_type)
LL, (LH, HL, HH) = coeffs2
fig = plt.figure(figsize=(11, 11))
for i, a in enumerate([LL, LH, HL, HH]):
ax = fig.add_subplot(2, 2, i + 1)
ax.imshow(a, interpolation="nearest", cmap=plt.cm.gray)
ax.set_title(titles[i], fontsize=10)
ax.set_xticks([])
ax.set_yticks([])
fig.tight_layout()
plt.show()
return
def vectorLBP(segmentation_type, color_type):
# IMAGES FOR TRAINING
# folders of trained and tested images based on chosen argument of user
if (color_type == "b"): # blue images
path_training = './dataset/blueTrain/*'
path_testing = './dataset/blueTest/*'
elif (color_type == "g"): # green images
path_training = './dataset/greenTrain/*'
path_testing = './dataset/greenTest/*'
elif (color_type == "r"): # red images
path_training = './dataset/redTrain/*'
path_testing = './dataset/redTest/*'
elif (color_type == "all"): # all images
path_training = './dataset/allTrain/*'
path_testing = './dataset/allTest/*'
# csv files for trained images
f = open("./csvFiles/LBPtrained.csv", "w+")
g = open("./csvFiles/LBPtrainedResult.csv", "w+")
for file in glob.glob(path_training):
file_substr = file.split('/')[-1] # get name of processed file
print(file_substr)
img = cv2.imread(file, 0) # uint8 image in grayscale
img = cv2.normalize(img, None, 0, 255,
cv2.NORM_MINMAX) # normalize image
# decide about type of segmentation
if (segmentation_type == "otsu"):
segmented_img = processedSegmentation.imgSegmentation(img)
elif (segmentation_type == "gauss"):
segmented_img = processedSegmentation.adaptiveSegmentationGaussian(
img)
elif (segmentation_type == "mean"):
segmented_img = processedSegmentation.adaptiveSegmentationMean(img)
# save, load segmented image and get their properties
cv2.imwrite('./processedImg/segmented_img.jpg', segmented_img)
img = cv2.imread('./processedImg/segmented_img.jpg')
height, width, channel = img.shape
img_gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
lbp_image = np.zeros((height, width, 3), np.uint8)
# processing LBP algorithm
lbp_values = []
for i in range(0, height):
for j in range(0, width):
lbp_image[i, j] = processLBP(img_gray, i, j, lbp_values)
hist_lbp = cv2.calcHist([lbp_image], [0], None, [256], [0, 256])
histogram_list = list()
histogram_list = hist_lbp
sum_hist1 = 0
sum_hist2 = 0
sum_hist3 = 0
sum_hist4 = 0
# get values from LBP histogram
for i in range(0, 64):
hist_value = histogram_list[i]
hist_value_n = str(hist_value[0])
sum_hist1 = sum_hist1 + float(hist_value_n)
for i in range(64, 128):
hist_value = histogram_list[i]
hist_value_n = str(hist_value[0])
sum_hist2 = sum_hist2 + float(hist_value_n)
for i in range(128, 192):
hist_value = histogram_list[i]
hist_value_n = str(hist_value[0])
sum_hist3 = sum_hist3 + float(hist_value_n)
for i in range(192, 256):
hist_value = histogram_list[i]
hist_value_n = str(hist_value[0])
sum_hist4 = sum_hist4 + float(hist_value_n)
# transform to more proper range for classification
sum_hist1_div = sum_hist1 / 1000000
sum_hist2_div = sum_hist2 / 1000000
sum_hist3_div = sum_hist3 / 1000000
sum_hist4_div = sum_hist4 / 1000000
img = cv2.imread('./processedImg/segmented_img.jpg',
0) # load input grayscale img
image = img_as_ubyte(img)
bins = np.array([
0, 16, 32, 48, 64, 80, 96, 112, 128, 144, 160, 176, 192, 208, 224,
240, 255
]) #16-bit
inds = np.digitize(image, bins)
max_value = inds.max() + 1
matrix_coocurrence = greycomatrix(inds, [1], [0],
levels=max_value,
normed=False,
symmetric=False)
# GLCM matrix characteristics - contrast, homogeinity, energy, correlation
contrast = greycoprops(matrix_coocurrence, 'contrast')
print("Contrast:")
print(contrast)
contrast_float = float(contrast)
contrast_str = contrast_float / 10 # transform to more proper range for classification
homogeneity = greycoprops(matrix_coocurrence, 'homogeneity')
print("Homogeneity:")
print(homogeneity)
homogeneity_float = float(homogeneity)
homogeneity_str = homogeneity_float / 10
energy = greycoprops(matrix_coocurrence, 'energy')
print("Energy:")
print(energy)
energy_float = float(energy)
energy_str = energy_float / 10
correlation = greycoprops(matrix_coocurrence, 'correlation')
print("Correlation:")
print(correlation)
correlation_float = float(correlation)
correlation_str = correlation_float / 10
saved_str = str(sum_hist1_div) + ", " + str(
sum_hist2_div) + ", " + str(sum_hist3_div) + ", " + str(
sum_hist4_div) + ", " + str(contrast_str) + ", " + str(
homogeneity_str) + ", " + str(energy_str) + ", " + str(
correlation_str) + "\n"
f.write(saved_str) # write vector to file
# save known result of image based on its file name
if ("fake" in file_substr):
print("This is known FAKE image for training")
g.write("0\n")
elif ("live" in file_substr):
print("This is known LIVE image for training")
g.write("1\n")
print()
# IMAGES FOR TESTING
h = open("./csvFiles/LBPtested.csv", "w+") # csv file for tested images
for file in glob.glob(path_testing):
file_substr = file.split('/')[-1] # get name of processed file
print(file_substr)
img = cv2.imread(file, 0) # uint8 image in grayscale
img = cv2.normalize(img, None, 0, 255,
cv2.NORM_MINMAX) # normalized image
# decide about type of segmentation
if (segmentation_type == "otsu"):
segmented_img = processedSegmentation.imgSegmentation(img)
elif (segmentation_type == "gauss"):
segmented_img = processedSegmentation.adaptiveSegmentationGaussian(
img)
elif (segmentation_type == "mean"):
segmented_img = processedSegmentation.adaptiveSegmentationMean(img)
# save, load segmented image and get their properties
cv2.imwrite('./processedImg/segmented_img.jpg', segmented_img)
img = cv2.imread('./processedImg/segmented_img.jpg')
height, width, channel = img.shape
img_gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
lbp_image = np.zeros((height, width, 3), np.uint8)
# processing LBP algorithm
lbp_values = []
for i in range(0, height):
for j in range(0, width):
lbp_image[i, j] = processLBP(img_gray, i, j, lbp_values)
hist_lbp = cv2.calcHist([lbp_image], [0], None, [256], [0, 256])
histogram_list = list()
histogram_list = hist_lbp
sum_hist1 = 0
sum_hist2 = 0
sum_hist3 = 0
sum_hist4 = 0
# get values from LBP histogram
for i in range(0, 64):
hist_value = histogram_list[i]
hist_value_n = str(hist_value[0])
sum_hist1 = sum_hist1 + float(hist_value_n)
for i in range(64, 128):
hist_value = histogram_list[i]
hist_value_n = str(hist_value[0])
sum_hist2 = sum_hist2 + float(hist_value_n)
for i in range(128, 192):
hist_value = histogram_list[i]
hist_value_n = str(hist_value[0])
sum_hist3 = sum_hist3 + float(hist_value_n)
for i in range(192, 256):
hist_value = histogram_list[i]
hist_value_n = str(hist_value[0])
sum_hist4 = sum_hist4 + float(hist_value_n)
# transform to more proper range for classification
sum_hist1_div = sum_hist1 / 1000000
sum_hist2_div = sum_hist2 / 1000000
sum_hist3_div = sum_hist3 / 1000000
sum_hist4_div = sum_hist4 / 1000000
img = cv2.imread('./processedImg/segmented_img.jpg', 0)
image = img_as_ubyte(img)
bins = np.array([
0, 16, 32, 48, 64, 80, 96, 112, 128, 144, 160, 176, 192, 208, 224,
240, 255
]) #16-bit
inds = np.digitize(image, bins)
max_value = inds.max() + 1
matrix_coocurrence = greycomatrix(inds, [1], [0],
levels=max_value,
normed=False,
symmetric=False)
# GLCM matrix characteristics - contrast, homogeinity, energy, correlation
contrast = greycoprops(matrix_coocurrence, 'contrast')
print("Contrast:")
print(contrast)
contrast_float = float(contrast)
contrast_str = contrast_float / 10 # transform to more proper range for classification
homogeneity = greycoprops(matrix_coocurrence, 'homogeneity')
print("Homogeneity:")
print(homogeneity)
homogeneity_float = float(homogeneity)
homogeneity_str = homogeneity_float / 10
energy = greycoprops(matrix_coocurrence, 'energy')
print("Energy:")
print(energy)
energy_float = float(energy)
energy_str = energy_float / 10
correlation = greycoprops(matrix_coocurrence, 'correlation')
print("Correlation:")
print(correlation)
correlation_float = float(correlation)
correlation_str = correlation_float / 10
saved_str = file_substr + ", " + str(sum_hist1_div) + ", " + str(
sum_hist2_div) + ", " + str(sum_hist3_div) + ", " + str(
sum_hist4_div) + ", " + str(contrast_str) + ", " + str(
homogeneity_str) + ", " + str(energy_str) + ", " + str(
correlation_str) + "\n"
print("This image will be used for liveness prediction")
h.write(saved_str) # write vector to file
print()
# properly close all csv files
f.close()
g.close()
h.close()
return
def vectorWavelet(segmentation_type, color_type, wavelet_type):
# IMAGES FOR TRAINING
# csv files for trained images
f = open("./csvFiles/WaveletTrained.csv", "w+")
g = open("./csvFiles/WaveletTrainedResult.csv", "w+")
# folders of trained and tested images based on chosen argument of user
if (color_type == "b"): # blue images
path_training = './dataset/blueTrain/*'
path_testing = './dataset/blueTest/*'
elif (color_type == "g"): # green images
path_training = './dataset/greenTrain/*'
path_testing = './dataset/greenTest/*'
elif (color_type == "r"): # red images
path_training = './dataset/redTrain/*'
path_testing = './dataset/redTest/*'
elif (color_type == "all"): # all images
path_training = './dataset/allTrain/*'
path_testing = './dataset/allTest/*'
for file in glob.glob(path_training):
img = cv2.imread(file, 0) # uint8 image in grayscale
img = cv2.normalize(img, None, 0, 255,
cv2.NORM_MINMAX) # normalize image
if (segmentation_type == "otsu"):
segmented_img = processedSegmentation.imgSegmentation(img)
elif (segmentation_type == "gauss"):
segmented_img = processedSegmentation.adaptiveSegmentationGaussian(
img)
elif (segmentation_type == "mean"):
segmented_img = processedSegmentation.adaptiveSegmentationMean(img)
cv2.imwrite('./processedImg/segmented_img.jpg', segmented_img)
image = cv2.imread('./processedImg/segmented_img.jpg')
image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
# Convert to float32 for more resolution for use with pywt
image = np.float32(image)
image /= 255
coeffs2 = pywt.dwt2(image, wavelet_type)
LL, (LH, HL, HH) = coeffs2
# gaining LH, HL and HH image
plt.imshow(LH, interpolation="nearest", cmap=plt.cm.gray)
plt.axis('off')
plt.savefig('./processedImg/LHimg.png',
transparent=True,
bbox_inches='tight',
pad_inches=0)
plt.imshow(HL, interpolation="nearest", cmap=plt.cm.gray)
plt.axis('off')
plt.savefig('./processedImg/HLimg.png',
transparent=True,
bbox_inches='tight',
pad_inches=0)
plt.imshow(HH, interpolation="nearest", cmap=plt.cm.gray)
plt.axis('off')
plt.savefig('./processedImg/HHimg.png',
transparent=True,
bbox_inches='tight',
pad_inches=0)
# PROCESS LH IMAGE - Horizontal detail
img = cv2.imread('./processedImg/LHimg.png', 0)
image = img_as_ubyte(img)
bins = np.array([
0, 16, 32, 48, 64, 80, 96, 112, 128, 144, 160, 176, 192, 208, 224,
240, 255
]) #16-bit
inds = np.digitize(image, bins)
max_value = inds.max() + 1
matrix_coocurrence = greycomatrix(inds, [1], [0],
levels=max_value,
normed=False,
symmetric=False)
# GLCM matrix characteristics - contrast, homogeinity, energy, correlation
contrast2 = greycoprops(matrix_coocurrence, 'contrast')
print("Contrast:")
print(contrast2)
contrast_float2 = float(contrast2)
contrast_str2 = contrast_float2 / 10
homogeneity2 = greycoprops(matrix_coocurrence, 'homogeneity')
print("Homogeneity:")
print(homogeneity2)
homogeneity_float2 = float(homogeneity2)
homogeneity_str2 = homogeneity_float2 / 10
energy2 = greycoprops(matrix_coocurrence, 'energy')
print("Energy:")
print(energy2)
energy_float2 = float(energy2)
energy_str2 = energy_float2 / 10
correlation2 = greycoprops(matrix_coocurrence, 'correlation')
print("Correlation:")
print(correlation2)
correlation_float2 = float(correlation2)
correlation_str2 = correlation_float2 / 10
# PROCESS HL IMAGE - Vertical detail
img = cv2.imread('./processedImg/HLimg.png', 0)
image = img_as_ubyte(img)
bins = np.array([
0, 16, 32, 48, 64, 80, 96, 112, 128, 144, 160, 176, 192, 208, 224,
240, 255
]) #16-bit
inds = np.digitize(image, bins)
max_value = inds.max() + 1
matrix_coocurrence = greycomatrix(inds, [1], [0],
levels=max_value,
normed=False,
symmetric=False)
contrast3 = greycoprops(matrix_coocurrence, 'contrast')
print("Contrast:")
print(contrast3)
contrast_float3 = float(contrast3)
contrast_str3 = contrast_float3 / 10
homogeneity3 = greycoprops(matrix_coocurrence, 'homogeneity')
print("Homogeneity:")
print(homogeneity3)
homogeneity_float3 = float(homogeneity3)
homogeneity_str3 = homogeneity_float3 / 10
energy3 = greycoprops(matrix_coocurrence, 'energy')
print("Energy:")
print(energy3)
energy_float3 = float(energy3)
energy_str3 = energy_float3 / 10
correlation3 = greycoprops(matrix_coocurrence, 'correlation')
print("Correlation:")
print(correlation3)
correlation_float3 = float(correlation3)
correlation_str3 = correlation_float3 / 10
# PROCESS HH IMAGE - Diagonal detail
img = cv2.imread('./processedImg/HHimg.png', 0)
image = img_as_ubyte(img)
bins = np.array([
0, 16, 32, 48, 64, 80, 96, 112, 128, 144, 160, 176, 192, 208, 224,
240, 255
]) #16-bit
inds = np.digitize(image, bins)
max_value = inds.max() + 1
matrix_coocurrence = greycomatrix(inds, [1], [0],
levels=max_value,
normed=False,
symmetric=False)
contrast4 = greycoprops(matrix_coocurrence, 'contrast')
print("Contrast:")
print(contrast4)
contrast_float4 = float(contrast4)
contrast_str4 = contrast_float4 / 10
homogeneity4 = greycoprops(matrix_coocurrence, 'homogeneity')
print("Homogeneity:")
print(homogeneity4)
homogeneity_float4 = float(homogeneity4)
homogeneity_str4 = homogeneity_float4 / 10
energy4 = greycoprops(matrix_coocurrence, 'energy')
print("Energy:")
print(energy4)
energy_float4 = float(energy4)
energy_str4 = energy_float4 / 10
correlation4 = greycoprops(matrix_coocurrence, 'correlation')
print("Correlation:")
print(correlation4)
correlation_float4 = float(correlation4)
correlation_str4 = correlation_float4 / 10
saved_str = (str(contrast_str2) + ", " + str(homogeneity_str2) + ", " +
str(energy_str2) + ", " + str(correlation_str2) + ", " +
str(contrast_str3) + ", " + str(homogeneity_str3) + ", " +
str(energy_str3) + ", " + str(correlation_str3) + ", " +
str(contrast_str4) + ", " + str(homogeneity_str4) + ", " +
str(energy_str4) + ", " + str(correlation_str4) + "\n")
print(saved_str)
file_substr = file.split('/')[-1]
f.write(saved_str) # write vector to file
# save known result of image based on its file name
if ("fake" in file_substr):
print("This is FAKE image.")
g.write("0\n")
elif ("live" in file_substr):
print("This is LIVE image.")
g.write("1\n")
# IMAGES FOR TESTING
h = open("./csvFiles/WaveletTested.csv",
"w+") # csv file for tested images
for file in glob.glob(path_testing):
img = cv2.imread(file, 0) # uint8 image in grayscale
img = cv2.normalize(img, None, 0, 255,
cv2.NORM_MINMAX) # normalize image
if (segmentation_type == "otsu"):
segmented_img = processedSegmentation.imgSegmentation(img)
elif (segmentation_type == "gauss"):
segmented_img = processedSegmentation.adaptiveSegmentationGaussian(
img)
elif (segmentation_type == "mean"):
segmented_img = processedSegmentation.adaptiveSegmentationMean(img)
cv2.imwrite('./processedImg/segmented_img.jpg', segmented_img)
image = cv2.imread('./processedImg/segmented_img.jpg')
image = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY)
# Convert to float for more resolution for use with pywt
image = np.float32(image)
image /= 255
coeffs2 = pywt.dwt2(image, wavelet_type)
LL, (LH, HL, HH) = coeffs2
plt.imshow(LH, interpolation="nearest", cmap=plt.cm.gray)
plt.axis('off')
plt.savefig('./processedImg/LHimg.png',
transparent=True,
bbox_inches='tight',
pad_inches=0)
plt.imshow(HL, interpolation="nearest", cmap=plt.cm.gray)
plt.axis('off')
plt.savefig('./processedImg/HLimg.png',
transparent=True,
bbox_inches='tight',
pad_inches=0)
plt.imshow(HH, interpolation="nearest", cmap=plt.cm.gray)
plt.axis('off')
plt.savefig('./processedImg/HHimg.png',
transparent=True,
bbox_inches='tight',
pad_inches=0)
# PROCESS LH IMAGE
img = cv2.imread('./processedImg/LHimg.png', 0)
image = img_as_ubyte(img)
bins = np.array([
0, 16, 32, 48, 64, 80, 96, 112, 128, 144, 160, 176, 192, 208, 224,
240, 255
]) #16-bit
inds = np.digitize(image, bins)
max_value = inds.max() + 1
matrix_coocurrence = greycomatrix(inds, [1], [0],
levels=max_value,
normed=False,
symmetric=False)
# GLCM matrix characteristics - contrast, homogeinity, energy, correlation
contrast2 = greycoprops(matrix_coocurrence, 'contrast')
print("Contrast:")
print(contrast2)
contrast_float2 = float(contrast2)
contrast_str2 = contrast_float2 / 10
homogeneity2 = greycoprops(matrix_coocurrence, 'homogeneity')
print("Homogeneity:")
print(homogeneity2)
homogeneity_float2 = float(homogeneity2)
homogeneity_str2 = homogeneity_float2 / 10
energy2 = greycoprops(matrix_coocurrence, 'energy')
print("Energy:")
print(energy2)
energy_float2 = float(energy2)
energy_str2 = energy_float2 / 10
correlation2 = greycoprops(matrix_coocurrence, 'correlation')
print("Correlation:")
print(correlation2)
correlation_float2 = float(correlation2)
correlation_str2 = correlation_float2 / 10
# PROCESS HL IMAGE
img = cv2.imread('./processedImg/HLimg.png', 0)
image = img_as_ubyte(img)
bins = np.array([
0, 16, 32, 48, 64, 80, 96, 112, 128, 144, 160, 176, 192, 208, 224,
240, 255
]) #16-bit
inds = np.digitize(image, bins)
max_value = inds.max() + 1
matrix_coocurrence = greycomatrix(inds, [1], [0],
levels=max_value,
normed=False,
symmetric=False)
contrast3 = greycoprops(matrix_coocurrence, 'contrast')
print("Contrast:")
print(contrast3)
contrast_float3 = float(contrast3)
contrast_str3 = contrast_float3 / 10
homogeneity3 = greycoprops(matrix_coocurrence, 'homogeneity')
print("Homogeneity:")
print(homogeneity3)
homogeneity_float3 = float(homogeneity3)
homogeneity_str3 = homogeneity_float3 / 10
energy3 = greycoprops(matrix_coocurrence, 'energy')
print("Energy:")
print(energy3)
energy_float3 = float(energy3)
energy_str3 = energy_float3 / 10
correlation3 = greycoprops(matrix_coocurrence, 'correlation')
print("Correlation:")
print(correlation3)
correlation_float3 = float(correlation3)
correlation_str3 = correlation_float3 / 10
# PROCESS HH IMAGE
img = cv2.imread('./processedImg/HHimg.png', 0)
image = img_as_ubyte(img)
bins = np.array([
0, 16, 32, 48, 64, 80, 96, 112, 128, 144, 160, 176, 192, 208, 224,
240, 255
]) #16-bit
inds = np.digitize(image, bins)
max_value = inds.max() + 1
matrix_coocurrence = greycomatrix(inds, [1], [0],
levels=max_value,
normed=False,
symmetric=False)
contrast4 = greycoprops(matrix_coocurrence, 'contrast')
print("Contrast:")
print(contrast4)
contrast_float4 = float(contrast4)
contrast_str4 = contrast_float4 / 10
homogeneity4 = greycoprops(matrix_coocurrence, 'homogeneity')
print("Homogeneity:")
print(homogeneity4)
homogeneity_float4 = float(homogeneity4)
homogeneity_str4 = homogeneity_float4 / 10
energy4 = greycoprops(matrix_coocurrence, 'energy')
print("Energy:")
print(energy4)
energy_float4 = float(energy4)
energy_str4 = energy_float4 / 10
correlation4 = greycoprops(matrix_coocurrence, 'correlation')
print("Correlation:")
print(correlation4)
correlation_float4 = float(correlation4)
correlation_str4 = correlation_float4 / 10
file_substr = file.split('/')[-1]
saved_str = (file_substr + ", " + str(contrast_str2) + ", " +
str(homogeneity_str2) + ", " + str(energy_str2) + ", " +
str(correlation_str2) + ", " + str(contrast_str3) + ", " +
str(homogeneity_str3) + ", " + str(energy_str3) + ", " +
str(correlation_str3) + ", " + str(contrast_str4) + ", " +
str(homogeneity_str4) + ", " + str(energy_str4) + ", " +
str(correlation_str4) + "\n")
print(saved_str)
h.write(saved_str) # write vector to file
# properly close all csv files
f.close()
g.close()
h.close()
return