def my_canny(img, fn = None, sigma=6, with_mask=False, save=False, show=False):
height = img.shape[0]
width = img.shape[1]
if with_mask:
import numpy as np
mymask = np.zeros((height, width),'uint8')
y1, x1 = 200, 150
y2, x2 = 500, 350
mymask[y1: y2, x1: x2] = 1
ret = canny(img, sigma=sigma, mask=mymask)
else:
ret = canny(img, sigma)
if show:
from src.utils.io import showimage_pil
showimage_pil(255*ret.astype('uint8'))
if save:
from src.utils.io import saveimage_pil
if with_mask:
feature = '_sigma' + str(sigma) + '_mask'
else:
feature = '_sigma' + str(sigma)
saveimage_pil(255*ret.astype('uint8'), fn+feature+'.jpg',show=False)
return ret
def Harris_Corner(arr, show=False, save=False, fn=None):
gray = np.float32(arr)
dst = cv2.cornerHarris(gray, 2, 3, 0.04)
dst = cv2.dilate(dst, None)
ret, dst = cv2.threshold(dst, 0.01 * dst.max(), 255, 0)
dst = np.uint8(dst)
ret, labels, stats, centroids = cv2.connectedComponentsWithStats(dst)
criteria = (cv2.TERM_CRITERIA_EPS + cv2.TERM_CRITERIA_MAX_ITER, 100, 0.001)
corners = cv2.cornerSubPix(gray, np.float32(centroids), (5, 5), (-1, -1), criteria)
res = np.hstack((centroids, corners))
res = np.int0(res)
arr[res[:, 1], res[:, 0]] = 255
if show is True:
showimage_pil(arr)
if save is True:
if fn is None:
return
else:
saveimage_pil(arr, fn)
def HOG(arr, show=False, save=False, fn=None):
from src.utils.util import normalize_array
from skimage.feature import hog
fd, hog_image = hog(
arr, orientations=8, pixels_per_cell=(16, 16), cells_per_block=(1, 1), visualise=True, normalise=True
)
if show is True:
# from skimage import exposure
# hog_image_rescaled = exposure.rescale_intensity(hog_image, in_range=(0, 1))
# showimage_pil(hog_image_rescaled)
hog_image = normalize_array(hog_image, low=0.2, high=1)
showimage_pil(hog_image)
if save is True:
if fn is None:
print "filename in src.utils.features.HOG is None"
return hog_image
# from skimage import exposure
# hog_image_rescaled = exposure.rescale_intensity(hog_image, in_range=(0, 1))
# saveimage_pil(hog_image_rescaled)
saveimage_pil(hog_image)
return fd
# fn = '/Users/ruhansa/Desktop/train/positive/flat/4/527.jpg'
img = Image.open(fn)
img.show()
cur_arr_3 = np.array(img.getdata(), dtype=np.uint8).reshape(img.size[1], img.size[0], -1)
cur_arr = cur_arr_3[:, :, 0]
from src.utils import preprocessing
import cv2
from src.utils.canny import decide_sigma
for sigma in [5]:
for neighbor in [30]:
blur = cv2.bilateralFilter(cur_arr.copy(), neighbor, sigma, sigma)
showimage_pil(blur)
cur_arr = blur
sigma = 0.0
r = 1
threshold = 0.025
while r > threshold:
print "sigma: " + str(sigma)
ret = my_canny(cur_arr, sigma=sigma, save=False, show=False)
s, r = decide_sigma(ret, ret.size, threshold=threshold,show=True)
if s is False:
sigma += 0.5
else:
if sigma == 0.0 and r < 0.01:
sharpen = preprocessing.sharpen(blur, sigma=2)
cur_arr= sharpen
showimage_pil(sharpen)
import os, sys, inspect from src.utils.io import showimage_pil, filename2arr tests_dir = os.path.dirname(os.path.abspath(inspect.getfile(inspect.currentframe()))) # script directory src/ src_dir = os.path.dirname(tests_dir) xray_dir = os.path.dirname(src_dir) # xray directory os.chdir(xray_dir) sys.path.append(src_dir) fn = xray_dir + "/data/LL/1.jpg" arr = filename2arr(fn) ratio_w = 300 / 450.0 ratio_h = 200 / 376.0 sw = arr.shape[1] * ratio_w sh = arr.shape[0] * ratio_h from src.utils.util import scale arr_new = scale(arr, sh, sw) showimage_pil(arr) showimage_pil(arr_new)
halves = non_max_suppression_merge(halves, overlapThresh=0)
if halves is not None:
# 1. detect ROI
bbox = get_huge_bounding_box(halves)
img1 = Image.fromarray(arr)
# fn = result_dir + '/' + testn + '_half_boxes.pkl'
# pickle.dump(halves, open(fn, "wb"))
img1=add_boxes(halves, img1)
img1.show()
img1.save(result_dir +'/' + testn +'_half_boxes.jpg')
# 2. detect edges
ROI = arr[bbox[1]: bbox[3], bbox[0]: bbox[2]]
edges, sigma, cur_ROI= best_edges(ROI, threshold=0.02)
showimage_pil(edges)
# ## 3. using hough_transform to find initial horizontal line
#
# from src.utils.hough import hough_horizontal
# fn = result_dir + '/' + testn + 'canny_hough'
# img2 = Image.fromarray(ROI)
# lines_init, _ = hough_horizontal(edges, fn, hough_line_len=30, line_gap=40, save=True, show=True, raw=img2, xdiff=0, ydiff=0)
#
#
# ## 4. Extend line using smaller sigma
# from src.utils.util import extend_line_canny_points
# from src.utils.canny import my_canny
# fn = fn = result_dir + '/' + testn + 'canny_extend'
# showimage_pil(cur_ROI)
xray_dir = os.path.dirname(src_dir) #xray directory
os.chdir(xray_dir)
sys.path.append(src_dir)
"""
###############################
set input/output path
###############################
"""
from src.utils import features
from src.utils import preprocessing
lns = ['L3']
test_n = 1
max_test=1
for ln in lns:
input_dir = xray_dir + '/data/train/' + str(ln)
for dirName, subdirList, fileList in os.walk(input_dir):
for filename in fileList:
parts = filename.split('.')
if parts[0] == '':
continue
if parts[1] == 'jpg' and test_n<= max_test:
test_n += 1
img = Image.open(dirName + '/'+ filename)
arr_3 = np.array(img.getdata(), dtype=np.uint8).reshape(img.size[1], img.size[0], 3)
arr = arr_3[:, :, 0]
# arr = preprocessing.normalize(arr)
# arr = preprocessing.sharpen(arr)
arr = preprocessing.deskew(arr)
showimage_pil(arr)
def drawMatches(img1, kp1, img2, kp2, matches):
"""
My own implementation of cv2.drawMatches as OpenCV 2.4.9
does not have this function available but it's supported in
OpenCV 3.0.0
This function takes in two images with their associated
keypoints, as well as a list of DMatch data structure (matches)
that contains which keypoints matched in which images.
An image will be produced where a montage is shown with
the first image followed by the second image beside it.
Keypoints are delineated with circles, while lines are connected
between matching keypoints.
img1,img2 - Grayscale images
kp1,kp2 - Detected list of keypoints through any of the OpenCV keypoint
detection algorithms
matches - A list of matches of corresponding keypoints through any
OpenCV keypoint matching algorithm
"""
# Create a new output image that concatenates the two images together
# (a.k.a) a montage
rows1 = img1.shape[0]
cols1 = img1.shape[1]
rows2 = img2.shape[0]
cols2 = img2.shape[1]
out = np.zeros((max([rows1, rows2]), cols1 + cols2, 3), dtype="uint8")
# Place the first image to the left
out[:rows1, :cols1, :] = np.dstack([img1, img1, img1])
# Place the next image to the right of it
out[:rows2, cols1 : cols1 + cols2, :] = np.dstack([img2, img2, img2])
# For each pair of points we have between both images
# draw circles, then connect a line between them
for mat in matches:
# Get the matching keypoints for each of the images
img1_idx = mat.queryIdx
img2_idx = mat.trainIdx
# x - columns
# y - rows
(x1, y1) = kp1[img1_idx].pt
(x2, y2) = kp2[img2_idx].pt
# Draw a small circle at both co-ordinates
# radius 4
# colour blue
# thickness = 1
cv2.circle(out, (int(x1), int(y1)), 4, (255, 0, 0), 1)
cv2.circle(out, (int(x2) + cols1, int(y2)), 4, (255, 0, 0), 1)
# Draw a line in between the two points
# thickness = 1
# colour blue
cv2.line(out, (int(x1), int(y1)), (int(x2) + cols1, int(y2)), (255, 0, 0), 1)
# Show the image
from src.utils.io import showimage_pil
showimage_pil(out)
