def main(args):
# Load the image
img_left_clr = read_image(args.image)
# Perform a some transforms on it for feature matching
tform = AffineTransform(scale=(0.8, 0.8), rotation=0.2, translation=(20, -10))
img_right_clr = warp(img_left_clr, tform.inverse, output_shape=img_left_clr.shape[:2])
img_right_clr = np.uint8(img_right_clr * 255)
# Convert to grayscale
img_left = cv2.cvtColor(img_left_clr, cv2.COLOR_BGR2GRAY)
img_right = cv2.cvtColor(img_right_clr, cv2.COLOR_RGB2GRAY)
print("Getting the features from the Harris Detector")
ftL = harris(img_left, sigma=3, threshold=0.01)
draw_corners(ftL, img_left_clr, 'corners_left')
ftR = harris(img_right, sigma=3, threshold=0.01)
draw_corners(ftR, img_right_clr, 'corners_right')
print(" -- Number of features (left): ", len(ftL))
print(" -- Number of features (right): ", len(ftR))
print("Finding the best matches between images")
with Timer(verbose=True) as t:
max_matches = min(len(ftL), len(ftR))
ptsL,ptsR = get_matches(ftL, ftR, [], [], img_left, img_right, max_matches, win_size=args.win_size)
print(" -- Number of matches = ", len(ptsL))
assert len(ptsL) == len(ptsR)
print("Performing RANSAC")
matches = ransac(ptsL,ptsR, img_left, img_right, args.max_iters, args.epsilon)
print(" -- Number of pruned matches = ", len(matches))
draw_matches(matches, img_left_clr, img_right_clr)
def get_matches(im1, im2): im1_corners, im2_corners = harris(im1, save=True, saveloc="lpoints.png"), harris(im2, save=True, saveloc="rpoints.png") im1_kps = harris_keypoints(im1, im1_corners) im2_kps = harris_keypoints(im2, im2_corners) return match(im1_kps, im2_kps)
def _run_detector(self, event=None):
try:
threshold = float(self._threshold.get())
sigma = float(self._sigma.get())
wwidth = int(self._wwidth.get())
wheight = int(self._wheight.get())
except (TypeError, ValueError):
raise
if self._img_show.img is None:
return
invert_img = self._invert.get()
images = harris(self._img_show.img,
threshold,
sigma,
wwidth,
wheight,
invert=invert_img)
mask, new_img = images[0], images[1]
if new_img is None:
return
new_img_tk = ImageTk.PhotoImage(new_img)
mimg = Image.new('L', (len(mask), len(mask[0])))
mimg.putdata([
255 * mask[i][j] for i in xrange(len(mask))
for j in xrange(len(mask[0]))
])
mimg = ImageTk.PhotoImage(mimg)
window = Tkinter.Toplevel()
window.image = new_img
label = Tkinter.Label(window, text='%s, %s' % (threshold, sigma))
mask_img = Tkinter.Label(window, image=mimg)
mask_img.img = mimg
img_frame = Tkinter.Frame(window)
img = scrolledcanvas(window, img_frame, highlightthickness=0)
img.new_img = new_img_tk
canvas_addimage(img, new_img_tk, new_img_tk.width(),
new_img_tk.height())
label.pack()
mask_img.pack()
img.pack(fill='both', expand=True)
img_frame.pack(fill='both', expand=True)
for image in images[2:]:
win = Tkinter.Toplevel()
win.image = image
iimg = ImageTk.PhotoImage(win.image)
iframe = Tkinter.Frame(win)
ilbl = scrolledcanvas(win, iframe, highlightthickness=0)
canvas_addimage(ilbl, iimg, *image.size)
ilbl.img = iimg
ilbl.pack(fill='both', expand=True)
iframe.pack(fill='both', expand=True)
def get_Srcextremes(self,ims,sa):
"""
find local extremas on pattern image
"""
# instantiate funtional class
hs = harris.harris()
hess = hessian.hessian()
cont = contrast.contrast()
coordinates = []
temp = {}
H = [0,1,2,3]
W = [0,1,2,3]
for i in range(4):
H[i] = len(ims[i][0])
W[i] = len(ims[i][0][0])
localArea = [0,1,2]
# get the unstable and low contrast pixel
hs_points = hs.corner(sa)
hess_points = hess.Srcedgedetect(sa)
low_contrast = cont.lowcontrast(sa)
# compute the pixels which are not situable for pixel matching
bad_points = list(set(hs_points) | set(hess_points) | set(low_contrast))
bad = dict.fromkeys(bad_points, 0)
for m in range(4):
for n in range(1,3):
for i in range(16,H[m]-16):
for j in range(16,W[m]-16):
if ((i*2**m,j*2**m) in bad.keys())==False :
# compare local pixel with its 26 neighbour
currentPixel = ims[m][n][i][j]
localArea[0] = ims[m][n-1][i-1:i+2,j-1:j+2]
localArea[1] = ims[m][n][i-1:i+2,j-1:j+2]
localArea[2] = ims[m][n+1][i-1:i+2,j-1:j+2]
Area = numpy.array(localArea)
maxLocal = numpy.array(Area).max()
minLocal = numpy.array(Area).min()
if (currentPixel == maxLocal) or (currentPixel == minLocal):
if ((i*2**m,j*2**m) in temp.keys()) == False:
coordinates.append([int(i*2**m),int(j*2**m)])
temp[(i*2**m,j*2**m)] = [i*2**m,j*2**m]
return coordinates
def get_Srcextremes(self,ims,sa):
"""
find local extremas on pattern image
"""
# instantiate funtional class
hs = harris.harris()
hess = hessian.hessian()
cont = contrast.contrast()
coordinates = []
temp = {}
H = [0,1,2,3]
W = [0,1,2,3]
for i in range(4):
H[i] = len(ims[i][0])
W[i] = len(ims[i][0][0])
localArea = [0,1,2]
# get the unstable and low contrast pixel
hs_points = hs.corner(sa)
hess_points = hess.Srcedgedetect(sa)
low_contrast = cont.lowcontrast(sa)
# compute the pixels which are not situable for pixel matching
bad_points = list(set(hs_points) | set(hess_points) | set(low_contrast))
bad = dict.fromkeys(bad_points, 0)
for m in range(4):
for n in range(1,3):
for i in range(16,H[m]-16):
for j in range(16,W[m]-16):
if bad.has_key((i*2**m,j*2**m))==False :
# compare local pixel with its 26 neighbour
currentPixel = ims[m][n][i][j]
localArea[0] = ims[m][n-1][i-1:i+2,j-1:j+2]
localArea[1] = ims[m][n][i-1:i+2,j-1:j+2]
localArea[2] = ims[m][n+1][i-1:i+2,j-1:j+2]
Area = numpy.array(localArea)
maxLocal = numpy.array(Area).max()
minLocal = numpy.array(Area).min()
if (currentPixel == maxLocal) or (currentPixel == minLocal):
if temp.has_key((i*2**m,j*2**m)) == False:
coordinates.append([int(i*2**m),int(j*2**m)])
temp[(i*2**m,j*2**m)] = [i*2**m,j*2**m]
return coordinates
def _run_detector(self, event=None):
try:
threshold = float(self._threshold.get())
sigma = float(self._sigma.get())
wwidth = int(self._wwidth.get())
wheight = int(self._wheight.get())
except (TypeError, ValueError):
raise
if self._img_show.img is None:
return
invert_img = self._invert.get()
images = harris(self._img_show.img, threshold, sigma, wwidth, wheight,
invert=invert_img)
mask, new_img = images[0], images[1]
if new_img is None:
return
new_img_tk = ImageTk.PhotoImage(new_img)
mimg = Image.new('L', (len(mask), len(mask[0])))
mimg.putdata([255 * mask[i][j]
for i in xrange(len(mask)) for j in xrange(len(mask[0]))])
mimg = ImageTk.PhotoImage(mimg)
window = Tkinter.Toplevel()
window.image = new_img
label = Tkinter.Label(window, text='%s, %s' % (threshold, sigma))
mask_img = Tkinter.Label(window, image=mimg)
mask_img.img = mimg
img_frame = Tkinter.Frame(window)
img = scrolledcanvas(window, img_frame, highlightthickness=0)
img.new_img = new_img_tk
canvas_addimage(img, new_img_tk,new_img_tk.width(),new_img_tk.height())
label.pack()
mask_img.pack()
img.pack(fill='both', expand=True)
img_frame.pack(fill='both', expand=True)
for image in images[2:]:
win = Tkinter.Toplevel()
win.image = image
iimg = ImageTk.PhotoImage(win.image)
iframe = Tkinter.Frame(win)
ilbl = scrolledcanvas(win, iframe, highlightthickness=0)
canvas_addimage(ilbl, iimg, *image.size)
ilbl.img = iimg
ilbl.pack(fill='both', expand=True)
iframe.pack(fill='both', expand=True)
def demo(imgfile0,imgfile1,rth0,rth1,mth):
# read image, convert to gray
#img = io.imread(sys.argv[1]);
img=io.imread("data/plane.bmp")
img0 = io.imread(imgfile0)
img1 = io.imread(imgfile1)
print("orig0 shape: "+str(img0.shape))
print("orig1 shape: "+str(img1.shape))
gry0 = rgb2gray(img0)
gry1 = rgb2gray(img1)
print("gray0 shape: "+str(gry0.shape))
print("gray1 shape: "+str(gry1.shape))
# get list of locations of interest points
feat0 = harris(gry0,gk(3,3,1),rth0)
feat1 = harris(gry1,gk(3,3,1),rth1)
print("number of features0: "+str(len(feat0)))
print("number of features1: "+str(len(feat1)))
# make sift descriptors
des0 = sift(gry0,feat0)
des1 = sift(gry1,feat1)
showFeatures(img0,feat0,des0)
showFeatures(img1,feat1,des1)
M = matchSIFT(des0,des1,40,mth)
img2 = drawMatches(img0,feat0,img1,feat1,M)
io.imshow( img2,vmin=0,vmax=255,cmap="gray")
io.show()
counter = counter + 1 #plt.figure(1) #plt.imshow(Ig) #plt.figure(2) #plt.imshow(Jg_interpolated) #plt.show() T_field = orientation_tensor(I, 17, 2.0, 17, 2.0) #plt.imshow(T_field[:,:,1,1], cmap="gray") #print(T_field[:,:,1,1].shape) #print(I.shape) #plt.show() harris_response = harris(T_field, 0.05) #harris_thresholded1 = harris_response < 500 #harris_thresholded2 = harris_response > -500 #harris_thresholded = harris_thresholded1 * harris_thresholded2 harris_thresholded = (harris_response < -15000) | (harris_response > 15000) #harris_thresholded = harris_thresholded * harris_response max_img = ndimage.filters.maximum_filter(harris_thresholded, size=1) [row, col] = np.nonzero(harris_thresholded == max_img) print(row) print(col) plt.imshow(max_img, cmap="gray") plt.show()
harris_kappa = 0.08
num_keypoints = 200
nonmaximum_supression_radius = 8
descriptor_radius = 9
match_lambda = 4
img = cv.imread('../data/000000.png')
img_gray = cv.cvtColor(img, cv.COLOR_BGR2GRAY)
""" Part 1 - Calculate Corner Response Functions """
# Shi - Tomasi
shi_tomasi_scores = shi_tomasi(img, corner_patch_size)
# assert (min(len(shi_tomasi_scores) == len(shi_tomasi_scores)))
# % Harris
harris_scores = harris(img, corner_patch_size, harris_kappa)
# assert (min(size(harris_scores) == size(harris_scores)))
plt.figure()
plt.subplot(2, 2, 1)
plt.imshow(img)
plt.subplot(2, 2, 2)
plt.imshow(img)
plt.subplot(2, 2, 3)
ax = plt.gca()
ax.imshow(shi_tomasi_scores)
plt.title('Shi-Tomasi Scores')
plt.subplot(2, 2, 4)
for i in range(data_set):
print i
img_1 = img[i]
img_2 = img[i+1]
img_1_cy = img_cy[i]
img_2_cy = img_cy[i+1]
### get gray scale
img_1_gray = cv2.cvtColor(img_1_cy, cv2.COLOR_BGR2GRAY)
img_2_gray = cv2.cvtColor(img_2_cy, cv2.COLOR_BGR2GRAY)
### harris
### points are(x, y) not (row, col)
print 'harris'
points_1 = harris.harris(img_1_gray)
points_2 = harris.harris(img_2_gray)
### draw dots
img_1_harris = harris.drawDots(img_1_cy, points_1)
img_2_harris = harris.drawDots(img_2_cy, points_2)
### get feature
print 'getting features...'
feature_1 = matching.descriptor(img_1_gray, points_1)
feature_2 = matching.descriptor(img_2_gray, points_2)
print 'matching...'
pairs = matching.find_pair(points_1, feature_1, points_2, feature_2)
print 'Got ', len(pairs[0]), ' pairs'
start = 0