pyfeats/demo/demo.py

import os
import numpy as np
import cv2
import matplotlib.pyplot as plt
from pyfeats import *

#%% Pre-processing
def padImage(im, pad=2, value=0):
    TDLU=[1, 1, 1, 1]  #top, down, left, right pad
    out = im.copy()
    for _ in range(pad):
        out = cv2.copyMakeBorder(out, TDLU[0], TDLU[1], TDLU[2], TDLU[3],\
                                     cv2.BORDER_CONSTANT, None, value)
    return out

# Load image
path = os.getcwd() + '/data/'
image_name ='ultrasound.bmp'
image = cv2.imread(path + image_name, cv2.IMREAD_GRAYSCALE)

plt.imshow(image, cmap='gray')
plt.title('Initial Image')
plt.show()

# Load mask
points_name = 'points.out'
points = np.loadtxt(path + points_name, delimiter=',')
points = np.array(points, np.int32)
mask = cv2.fillPoly(np.zeros(image.shape,np.double), [points.reshape((-1,1,2))], color=1).astype('i')
perimeter = cv2.polylines(np.zeros(image.shape), [points.reshape((-1,1,2))],isClosed=True,color=1,thickness=1).astype('i')      

plt.imshow(mask, cmap='gray')
plt.title('Initial Mask')
plt.show()  

# Fill image with zeros outside ROI
image = np.multiply(image.astype(np.double), mask).astype(np.uint8)
max_point_x = max(points[:,0])
max_point_y = max(points[:,1])
min_point_x = min(points[:,0])
min_point_y = min(points[:,1])  
a,b,c,d = min_point_y, (max_point_y+1), min_point_x, (max_point_x+1)
image, mask, perimeter = image[a:b,c:d], mask[a:b,c:d], perimeter[a:b,c:d]
image, mask, perimeter = padImage(image, 2, 0), padImage(mask, 2, 0), padImage(perimeter, 2, 0) # ALWAYS PAD IMAGE SO AS SHAPE PARAMETERS WORK

plt.imshow(image, cmap='gray')
plt.title('Image cropped to ROI')
plt.show()

plt.imshow(mask, cmap='gray')
plt.title('Mask of image cropped to ROI')
plt.show()

plt.imshow(perimeter, cmap='gray')
plt.title('Perimeter of image cropped to ROI')
plt.show()

#%% A1. Texture features
features = {}
features['A_FOS'] = fos(image, mask)
features['A_GLCM'] = glcm_features(image, ignore_zeros=True)
features['A_GLDS'] = glds_features(image, mask, Dx=[0,1,1,1], Dy=[1,1,0,-1])
features['A_NGTDM'] = ngtdm_features(image, mask, d=1)
features['A_SFM'] = sfm_features(image, mask, Lr=4, Lc=4)
features['A_LTE'] = lte_measures(image, mask, l=7)
features['A_FDTA'] = fdta(image, mask, s=3)
features['A_GLRLM'] = glrlm_features(image, mask, Ng=256)
features['A_FPS'] = fps(image, mask)
features['A_Shape_Parameters'] = shape_parameters(image, mask, perimeter, pixels_per_mm2=1)
features['A_HOS'] = hos_features(image, th=[135,140])
features['A_LBP'] = lbp_features(image, image, P=[8,16,24], R=[1,2,3])
features['A_GLSZM'] = glszm_features(image, mask)
plot_sinogram(image, image_name)

#%% B. Morphological features
features['B_Morphological_Grayscale_pdf'], features['B_Morphological_Grayscale_cdf'] = grayscale_morphology_features(image, N=30)
features['B_Morphological_Binary_L_pdf'], features['B_Morphological_Binary_M_pdf'], features['B_Morphological_Binary_H_pdf'], features['B_Morphological_Binary_L_cdf'], \
features['B_Morphological_Binary_M_cdf'], features['B_Morphological_Binary_H_cdf'] = multilevel_binary_morphology_features(image, mask, N=30, thresholds=[25,50])

plot_pdf_cdf(features['B_Morphological_Grayscale_pdf'], features['B_Morphological_Grayscale_cdf'], image_name)
plot_pdfs_cdfs(features['B_Morphological_Binary_L_pdf'], features['B_Morphological_Binary_M_pdf'], features['B_Morphological_Binary_H_pdf'], features['B_Morphological_Binary_L_cdf'], features['B_Morphological_Binary_M_cdf'], features['B_Morphological_Binary_H_cdf'])

#%% C. Histogram Based features
features['C_Histogram'] = histogram(image, mask, bins=32)
features['C_MultiregionHistogram'] = multiregion_histogram(image, mask, bins=32, num_eros=3, square_size=3)
features['C_Correlogram'] = correlogram(image, mask, bins_digitize=32, bins_hist=32, flatten=True)

plot_histogram(image, mask, bins=32, name=image_name)
plot_correlogram(image, mask, bins_digitize=32, bins_hist=32, name=image_name)

#%% D. Multi-Scale features
features['D_DWT'] = dwt_features(image, mask, wavelet='bior3.3', levels=3)
features['D_SWT'] = swt_features(image, mask, wavelet='bior3.3', levels=3)
features['D_WP'] = wp_features(image, mask, wavelet='coif1', maxlevel=3)
features['D_GT'] = gt_features(image, mask)
features['D_AMFM'] = amfm_features(image)

#%% E. Other
features['E_HOG'] = hog_features(image, ppc=8, cpb=3)
features['E_HuMoments'] = hu_moments(image)
features['E_TAS'] = tas_features(image)
features['E_ZernikesMoments'] = zernikes_moments(image, radius=9)

#%% Print
for x, y in features.items():
    if x.startswith("B") | x.startswith("C") | x.startswith("E"):
        continue
    print('-' * 50)
    print(x)
    print('-' * 50)
    if len(y[1]) == 1:
        print(y[1][0], '=', y[0])
    else:
        for i in range(len(y[1])):
           print(y[1][i], '=', y[0][i])