def main():
imgname1='imgs/porta1pb.pnm'
imgname2='imgs/larapiopb.pnm'
if len(argv)>2:
imgname1=argv[1]
imgname2=argv[2]
thresh_factor = 65
min_size = 10 #(minimum w and h to be not considered noise)
print 'Subtracting...'
subtract(imgname1, imgname2)
print 'Thresholding...'
threshold('subtracted_image.png', 65)
print 'Eroding...'
k=open('kernel', 'w')
k.write('''1 1 1 1 1
1 1 1 1 1
1 1 1 1 1''')
k.close()
erode('thresholded_image.png')
print 'Dilating...'
dilate('eroded_image.png')
#Dilating to merge pieces, then erode to get back to the original size
print 'Merging pieces...',
k=open('kernel', 'w')
for i in range(10):
k.write('1 1 1 1 1 1 1 1 1 1\n')
k.close()
dilate('dilated_image.png', outfile='result0.png')
print 'halfway done...'
erode('result0.png', outfile = 'result.png')
#Using opencv to calculate the bounding box to figure out what we found
img = cv2.imread('result.png', cv2.CV_LOAD_IMAGE_GRAYSCALE)
contours, hierarchy = cv2.findContours(img,cv2.RETR_LIST,cv2.CHAIN_APPROX_SIMPLE)
noise=0
for cnt in contours:
b_box = cv2.boundingRect(cnt)
w,h = b_box[2], b_box[3]
#print w,h
if w > min_size and h > min_size:
#Found something...
if w > h:
print 'Found a dog!'
else:
print 'Found an intruder!'
else:
noise+=1
print 'Found',str(noise),'noise(s).'
def get_key_points(inputpath, outputpath):
segmentpath = inputpath.split(".")[:-1][0] + "_seg." + inputpath.split(
".")[-1]
erodepath = inputpath.split(".")[:-1][0] + "_erod." + inputpath.split(
".")[-1]
contour(inputpath, segmentpath)
erode(inputpath, segmentpath, erodepath)
op = mark_keypoints(inputpath, outputpath, segmentpath)
os.system("rm -rf {} {}".format(segmentpath, erodepath))
return op
def extract_finger(filename):
"""Yet to be written"""
blk_sz_c = 24
blk_sz_o = 24
img = mpimg.imread(filename)
#convert to gray scale
if img.ndim == 3: #colour image
img = rgb2gray(img)
yt = 1
yb = img.shape[1]
xr = img.shape[0]
xl = 1
YA = 0
YB = 0
XA = 0
XB = 0
delta_test = 0
for x in range(55):
if np.where(img[x,:] < 200)[0].size < 8:
img[0:x, :] = 255
yt = x + 1
for x in range(224, img.shape[0]):
if np.where(img[x,:] < 200)[0].size < 3:
img[x - 17:img.shape[0],:] = 255
yb = x + 1
break
for y in range(199, img.shape[1]):
if np.where(img[:,y] < 200)[0].size < 1:
img[:,y:img.shape[1]] = 255
xr = y + 1
break
for y in range(75):
if np.where(img[:,y] < 200)[0].size < 1:
img[:,0:y] = 255
xl = y + 1
enhimg = img.copy()
[cimg1, o1, fimg, bwimg, eimg, enhimg] = fft_enhance_cubs(enhimg, blk_sz_o/4)
[cimg, oimg2, fimg, bwimg, eimg, enhimg] = fft_enhance_cubs(enhimg, blk_sz_o/2)
[cimg2, oimg, fimg, bwimg, eimg, enhimg] = fft_enhance_cubs(enhimg, blk_sz_o)
[newim, binim, mask, o_rel, orient_img] = testfin(enhimg) #testfin is from Dr. Peter Kovesi's code
[cimg, oimg2, fimg, bwimg, eimg, enhimg] = fft_enhance_cubs(img, -1)
[newim, binim, mask, o_rel, orient_img_m] = testfin(enhimg)
theta1 = 0.5
mask_t = mask.copy()
for y in range(19, mask.shape[0] - blk_sz_c*2 + 1):
for x in range(blk_sz_c, mask.shape[1] - blk_sz_c*2 + 1):
n_mask = 0
for yy in range(-1,2):
for xx in range(-1,2):
y_t = y + yy*blk_sz_c
x_t = x + xx*blk_sz_c
if y_t > 0 and x_t > 0 and (y_t != y or x_t != x) and mask[y_t - 1, x_t - 1] == 0:
n_mask = n_mask + 1
if n_mask == 0:
continue
if mask[y - 1, x - 1] == 0 or y > mask.shape[0] - 20 or y < yt or y > yb or x < xl or x > xr:
cimg2[np.ceil(y/float(blk_sz_c)) - 1, np.ceil(x/float(blk_sz_c)) - 1] = 255
mask_t[y - 1, x - 1] = 0
continue
for i in range(y, y + 2):
for j in range(x - 9, x + 10):
if i > 0 and j > 0 and i < mask.shape[0] and j < mask.shape[1] and mask[i - 1, j - 1] > 0:
continue
else:
cimg2[np.ceil(y/float(blk_sz_c)) - 1, np.ceil(x/float(blk_sz_c)) - 1] = 255
mask_t[y - 1, x - 1] = 0
break
mask = mask_t.copy()
cimg1[np.where(cimg1 > 0.9)[0]] = 255
cimg1[np.where(cimg > 0.9)[0]] = 255
cimg2[np.where(cimg2 > 0.875)[0]] = 255
#this loop takes 10 minutes
for y in range(1, mask.shape[0] - blk_sz_c*2 + 1):
for x in range(blk_sz_c, mask.shape[1] - blk_sz_c*2 + 1):
for i in range(y - 25, y + 26):
for j in range(x - 25, x + 26):
if i > 0 and j > 0 and i < mask.shape[0] and j < mask.shape[1] and mask[i - 1, j - 1] > 0:
continue
else:
cimg2[np.ceil(y/float(blk_sz_c)) - 1, np.ceil(x/float(blk_sz_c)) - 1] = 255
break
for y in range(1, mask.shape[0] - blk_sz_c*2 + 1):
for x in range(1, mask.shape[1] - blk_sz_c + 1):
if cimg[np.ceil(y/(float(blk_sz_c)/2)) - 1, np.ceil(x/(float(blk_sz_c)/2)) - 1] == 255:
cimg1[np.ceil(y/(float(blk_sz_c)/4)) - 1, np.ceil(x/(float(blk_sz_c)/4)) - 1] = 255
cimg1[np.ceil(y/(float(blk_sz_c)/4)), np.ceil(x/(float(blk_sz_c)/4)) - 1] = 255
cimg1[np.ceil(y/(float(blk_sz_c)/4)) - 1, np.ceil(x/(float(blk_sz_c)/4))] = 255
cimg1[np.ceil(y/(float(blk_sz_c)/4)), np.ceil(x/(float(blk_sz_c)/4))] = 255
for y in range(1, mask.shape[0] - blk_sz_c + 1):
for x in range(1, mask.shape[1] - blk_sz_c + 1):
if cimg2[np.ceil(y/float(blk_sz_c)) - 1, np.ceil(x/float(blk_sz_c)) - 1] == 255:
if np.ceil(x/(float(blk_sz_c)/2)) <= cimg.shape[1] -1:
if np.ceil(y/(float(blk_sz_c)/2)) <= cimg.shape[0] -1:
cimg[np.ceil(y/(float(blk_sz_c)/2)), np.ceil(x/(float(blk_sz_c)/2)) - 1] = 255
cimg[np.ceil(y/(float(blk_sz_c)/2)), np.ceil(x/(float(blk_sz_c)/2))] = 255
cimg[np.ceil(y/(float(blk_sz_c)/2)) - 1, np.ceil(x/(float(blk_sz_c)/2))] = 255
else:
xpad = np.ceil(y/(float(blk_sz_c)/2))- (cimg.shape[0] -1)
else:
ypad = np.ceil(x/(float(blk_sz_c)/2)) - (cimg.shape[1] -1)
cimg = np.hstack((np.vstack((cimg, 255*np.ones((xpad,cimg.shape[1])))),np.vstack((255*np.ones((cimg.shape[0],ypad)),255*np.ones((xpad,ypad))))))
cimg1[np.where(cimg1 < 0.51)[0]] = 255
cimg[np.where(cimg < 0.51)[0]] = 255
cimg2[np.where(cimg2 < 0.51)[0]] = 255
inv_binim = (binim == 0)
pdb.set_trace()
thinned = erode(inv_binim) #What is bwmorph?
mask_t = mask.copy()
if np.where(mask[124:150, 149:250] > 0)[0].size > 0 and np.where(mask[249:275, 149:250] > 0)[0].size > 0:
mask[149:250, 149:250] = 1
method = -1
core_path = 255
core_y = 0
core_x = 0
core_val = 0
lc = 0
o_img = np.sin(orient_img)
o_img[np.where(mask == 0)[0]] = 1
lower_t = 0.1
v = cimg.min(0)
y = cimg.argmin(0) + 1
dt1 = v.min(0)
x = v.argmin(0) + 1
delta1_y = y[x - 1]*float(blk_sz_c)/2
delta1_x = x*float(blk_sz_c)/2
v[x - 1] = 255
v[x] = 255
dt2 = v.min(0)
x = v.argmin(0) + 1
delta2_y = y[x - 1]*float(blk_sz_c)/2
delta2_x = x*float(blk_sz_c)/2
v[x - 1] = 255
v[x] = 255
dt3 = v.min(0)
x = v.argmin(0) + 1
delta3_y = y[x - 1]*float(blk_sz_c)/2
delta3_x = x*float(blk_sz_c)/2
db = 60
if dt1 < 1 and delta1_y + db < core_y and delta1_y > 15 or dt2 < 1 and delta2_y + db < core_y and delta2_y > 15 or dt3 < 1 and delta3_y + db < core_y \
and delta3_y > 15:
core_val = 255
for y in range(10, o_img.shape[0] - 9):
for x in range(10, o_img.shape[1] - 9):
s1 = 0
t = 10
if y < 50 and x > 250:
t = 11
if y > 38:
yt = 20
else:
yt = 5
if lc > 0.41 and (core_y + 60 < y):
break
if mask[y - 1, x - 1] == 0 or mask[max(y - t,1) - 1, x - 1] == 0 or mask[y - 1, min(x + t, o_img.shape[1]) - 1] == 0 \
or mask[y - 1, max(x - t, 1) - 1] == 0 or mask[max(y - t, 1) - 1, min(x + t, o_img.shape[1]) - 1] == 0 \
and mask[max(y - t, 1) - 1, max(x - t, 1) - 1] == 0 or o_img[y - 1, x - 1] < lc or o_img[y - 1, x - 1] < 0.1:
continue
if dt1 < 1 and delta_y + db < y and delta_y > 15 or dt2 < 1 and delta2_y + db < y and delta2_y > 15 or \
dt3 < 1 and delta3_y + db < y and delta3_y > 15:
continue
test_m = np.min(o_img[0:y - yt, (np.max((x - 10), 1) -1):np.min(x + 10, o_img.shape[1])])
if test_m.shape > 0 and np.min(test_m) >= 0.17:
continue
for a in range(y, y + 3):
for b in range(x, x + 2):
s1 = s1 + o_img[a - 1, b - 1]
s1 = float(s1)/6
s2 = []
i = 1
for a in range(y - 3, y):
for b in range(x, x + 2):
s2[i - 1] = o_img[a - 1, b - 1]
i = i + 1
if min(s2) < lower_t:
s2 = sum(s2)/6
else:
s2 = s1 #s2 was a list, s1 is a no., this statement makes s2 a number.
s3 = []
i = 1
for a in range(y, y + 3):
for b in range(x + 2, x + 4):
s3[i - 1] = o_img[a - 1, b - 1]
i = i + 1
if min(s3) < lower_t:
s3 = sum(s3)/float(6)
else:
s3 = s1
s4 = []
i = 1
for a in range(y, y + 3):
for b in range(x - 2, x):
s4[i - 1] = o_img[a - 1, b - 1]
i = i + 1
if min(s4) < lower_t:
s4 = sum(s4)/float(6)
else:
s4 = s1
s5 = []
i = 1
for a in range(y - 3, y):
for b in range(x - 2, x):
s5[i - 1] = o_img[a - 1, b - 1]
i = i + 1
if min(s5) < lower_t:
s5 = sum(s5)/float(6)
else:
s5 = s1
s6 = []
for a in range(y - 3, y):
for b in range(x + 2, x + 4):
s6[i - 1] = o_img[a - 1, b - 1]
i = i + 1
if min(s6) < lower_t:
s6 = sum(s6)/float(6)
else:
s6 = s1
if s1 - s2 > core_val:
core_val = s1 - s2
core_x = x
core_y = y
lc = o_img[y - 1, x - 1]
method = 1
if s1 - s3 > core_val:
core_val = s1 - s3
core_x = x
core_y = y
lc = o_img[y - 1, x - 1]
method = 2
if x < 300 and s1 - s4 > core_val:
core_val = s1 - s4
core_x = x
core_y = y
lc = o_img[y - 1, x - 1]
method = 3
if x < 300 and s1 - s5 > core_val:
core_val = s1 - s5
core_x = x
core_y = y
lc = o_img[y - 1, x - 1]
method = 4
if s1 - s6 > core_val:
core_val = s1 - s6
core_x = x
core_y = y
lc = o_img[y - 1, x - 1]
method = 5
#nested for loop ends here
pdb.set_trace() #to be deleted
yt = 0
if core_y > 37:
yt = 20
else:
yt = 5
#Printing is as desired by MATLAB code, reason for printing not known.
print "lc = ", lc
print "core_y = ", core_y
print "core_x = ", core_x
print "method = ", method
test_smooth = 100
if core_y > 0:
test_smooth = np.sum(np.sum(o_img[core_y - yt - 6:core_y - yt + 5, core_x - 6: core_x + 5],0))
if lc > 0.41 and (test_smooth < 109.5 and method != 2 or test_smooth < 100):
start_t = 0
core_val = float(1)/(core_val + 1)
print "core_x = ", core_x
print "core_y = ", core_y
#How to code line 404?
else:
core_x = 0
core_y = 0
core_val = 255
mask = mask_t.copy()
display_flag = 1
ocodes = []
orientations = []
radii = []
sample_intv = 5
path_len = 45
min_o_change = np.zeros([1, floor(float(path_len)/sample_intv)])
#to append more rows in min_o_change, we'll use np.vstack() function
minu_count = 1
minutiae = np.array([0,0,0,0,0,1])
# Bibek's code starts here
pdb.set_trace() #to be deleted
# loop through image and find minutiae, ignore certain pixels for border
for y in range(20, (img.shape[0] - 14 +1)):
for x in range(21, (img.shape[1] - 21 +1)):
if (thinned[(y -1), (x -1)] == 1):
# only continue if pixel is white
# calculate CN from Raymond Thai
CN = 0
sx = 0
sy = 0
for i in range(1, 9):
t1, x1, y1 = p(thinned, x, y, i) # nargout=3
t2, x2, y2 = p(thinned, x, y, i + 1) # nargout=3
CN = CN + abs(t1 - t2)
CN = CN / 2
if ((CN == 1) or (CN == 3)): #&& mask(y,x) > 0
skip = 0
for i in range(y - 5, (y + 5 +1)):
for j in range(x - 5, (x + 5 +1)):
if np.logical_and(np.logical_and(i > 0, j > 0), mask[(i -1), (j -1)]) == 0:
skip = 1
if skip == 1:
continue
t_a = np.array([])
c = 0
for e in range(y - 1, (y + 1 +1)):
for f in range(x - 1, (x + 1 +1)):
c = c + 1
np.insert(t_a,(c -1),orient_img_m[(e -1), (f -1)])
m_o = np.median(t_a) #orient_img_m(y,x); #oimg(ceil(y/blk_sz_o),ceil(x/blk_sz_o));
m_f = 0 #fimg(ceil(y/blk_sz_o),ceil(x/blk_sz_o));
angle = m_o
if CN == 3:
CN, prog, sx, sy, ang = test_bifurcation(thinned, x, y, m_o, core_x, core_y) # nargout=5
if prog < 3:
continue
if ang < pi:
m_o = mod(m_o + pi, np.dot(2, pi))
# if ~(ang < pi && min(abs(m_o - ang),2*pi - abs(m_o - ang)) > abs(pi-abs(m_o - ang)))
# m_o = mod(m_o+pi,2*pi);
# end
print 'y = ' + str(y) +'\n'
print 'x = ' + str(x) +'\n'
print 'm_o = ' + str(m_o) +'\n'
print 'ang = ' + str(ang) +'\n'
print 'sy = ' + str(sy) +'\n'
print 'sx = ' + str(sx) +'\n'
print np.min(abs(m_o - ang), np.dot(2, pi) - abs(m_o - ang))
else:
progress = 0
xx = x
yy = y
pao = - 1
pos = 0
while progress < 15 and xx > 1 and yy > 1 and yy < img.shape[0] and xx < img.shape[1] and pos > - 1:
pos = - 1
for g in range(1, 9):
ta, xa, ya = p(thinned, xx, yy, g) # nargout=3
tb, xb, yb = p(thinned, xx, yy, g + 1) # nargout=3
if (ta > tb) and pos == - 1 and g != pao:
pos = ta
if g < 5:
pao = 4 + g
else:
pao = mod(4 + g, 9) + 1
xx = xa
yy = ya
progress = progress + 1
if progress < 10:
continue
if mod(atan2(y - yy, xx - x), np.dot(2, pi)) > pi:
m_o = m_o + pi
minutiae[(minu_count -1), :] = np.array([x, y, CN, m_o, m_f, 1]).reshape(1, -1)
min_path_index[(minu_count -1), :] = np.array([sx, sy])
minu_count = minu_count + 1
#end of if loop if pixel white
# enf of for y
# end of for x
# Identify the ellipse enclosed by the largest rectangle spanned by the
# set of initially detected minutiaes (4/8/2006 - Paul Kwan)
min_x = np.min(minutiae[:, 0])
max_x = np.max(minutiae[:, 0])
min_y = np.min(minutiae[:, 1]) / 1.1
max_y = np.dot(np.max(minutiae[:, 1]), 1.1)
mid_x = (min_x + max_x) / 2
mid_y = (min_y + max_y) / 2
if (max_x - min_x) > (max_y - min_y): # major axis is on the x-axis
major_axis_len = max_x - min_x # Refer to figure
M_Y1 = (max_y - min_y) / 2
F1_Y1 = major_axis_len / 2
M_F1 = sqrt(F1_Y1 ** 2 - M_Y1 ** 2) # Use Pythagoras' theorem
F1_x = mid_x - M_F1
F1_y = mid_y
F2_x = mid_x + M_F1
F2_y = mid_y
else:
major_axis_len = max_y - min_y # Refer to figure
M_X1 = (max_x - min_x) / 2
F1_X1 = major_axis_len / 2
M_F1 = sqrt(F1_X1 ** 2 - M_X1 ** 2) # Use Pythagoras' theorem
F1_x = mid_x
F1_y = mid_y + M_F1
F2_x = mid_x
F2_y = mid_y - M_F1
clean_list = np.array([])
c_index = 1
thinned_old = thinned.copy()
minu_count = minu_count - 1
# Filter potential false endings by excluding those outside of the
# elliptical region
t_minutiae = np.array([])
t_minu_count = 1
t_mpi = np.array([])
pre_minu_count1 = minu_count
for i in range(1, (minu_count +1)):
X = minutiae[(i -1), 0]
Y = minutiae[(i -1), 1]
rc = 0
for y in range(max(Y - 2, 1), (min(Y + 2, binim.shape[0]) +1)):
if rc > 0:
break
for x in range(max(X - 2, 1), (min(X + 2, binim.shape[1]) +1)):
if mask[(y -1), (x -1)] == 0:
rc = rc + 1
break
if rc > 0:
continue
else:
np.insert(t_minutiae, (t_minu_count -1), minutiae[(i -1), :])
np.insert(t_mpi, (t_minu_count -1), min_path_index[(i -1), :])
t_minu_count = t_minu_count + 1
minutiae = t_minutiae.copy()
min_path_index = t_mpi.copy()
minu_count = minutiae.shape[0]
t_minu_count = 1
t_minutiae = np.array([])
pre_minu_count2 = minu_count
print 'pre_minu_count2 =' + str(pre_minu_count2) + '\n'
dist_m = dist2(minutiae[:, 0:2], minutiae[:, 0:2])
#dist_m(find(dist_m == 0)) == 1000;
dist_test = 49
#if minu_count > 50
# dist_test=81;
#end
for i in range(1, (minu_count +1)):
reject_flag = 0
t_a = minutiae[(i -1), 3]
P_x = minutiae[(i -1), 0]
P_y = minutiae[(i -1), 1]
for j in range(i + 1, (minu_count +1)):
if dist_m[(i -1), (j -1)] <= dist_test:
reject_flag = 1
if reject_flag == 0 and mask[(P_y -1), (P_x -1)] > 0:
reverse_p = 0
if min_path_index[(i -1), 0] == 0:
x = P_x
y = P_y
else:
x = min_path_index[(i -1), 0]
y = min_path_index[(i -1), 1]
iter_ = 0
p1x = P_x
p1y = P_y
p2x = P_x
p2y = P_y
minutiae_o_change = np.array([])
x1 = x
y1 = y
xs = x
ys = y
p_uv = np.array([(x / math.sqrt(x ** 2 + y ** 2)), (y / math.sqrt(x ** 2 + y ** 2))])
iter_ = 0
for m in range(1, (path_len +1)):
iter_ = iter_ + 1
cn = 0
for ii in range(1, 9):
t1, x_A, y_A = p(thinned, x1, y1, ii) # nargout=3
t2, x_B, y_B = p(thinned, x1, y1, ii + 1) # nargout=3
cn = cn + abs(t1 - t2)
cn = cn / 2
if (cn != 3 and cn != 4) or m == 1: #&& mask(x1,y1) > 0
for n in range(1, 9):
if reverse_p == 0 or iter_ > 1:
ta, xa, ya = p(thinned, x1, y1, n) # nargout=3
else:
ta, xa, ya = p(thinned, x1, y1, 9 - n) # nargout=3
if ta == 1 and (xa != p1x or ya != p1y) and (xa != x or ya != y):
p1x = x1
p1y = y1
x1 = xa
y1 = ya
break
if np.mod(iter_, sample_intv) == 0:
if xs == x1 and ys == y1:
minutiae_o_change = np.insert(minutiae_o_change, (iter_ / sample_intv -1), - 1)
else:
norm_s = math.sqrt((x1 - xs) ** 2 + (y1 - ys) ** 2)
xv = (x1 - xs) / norm_s
yv = (y1 - ys) / norm_s
ptx = x - xs
pty = y - ys
norm_p = math.sqrt((ptx) ** 2 + (pty) ** 2)
ptx = ptx / norm_p
pty = pty / norm_p
#inner product
i_prod = (np.dot(xv, ptx) + np.dot(yv, pty))
minutiae_o_change = np.insert(minutiae_o_change,(iter_ / sample_intv -1), math.acos(i_prod))
xs = x1
ys = y1
minutiae_o_change = np.insert(minutiae_o_change, 0, 1)
# [temp_o1, temp_r1, mx1, my1] = tico(binim, P_x, P_y, oimg, orient_img_m, o_rel, [12 24 36 48 60 72 84] , [14 18 24 28 30 34 28] , 1, minutiae(i,4), blk_sz_o);
temp_o1, temp_r1, mx1, my1 = tico(binim, P_x, P_y, oimg, orient_img_m, o_rel, np.array([12, 24, 36, 48, 60, 72, 84]), np.array([14, 18, 24, 28, 30, 34, 28]), 0, - 1, blk_sz_o) # nargout=4
t_minutiae = np.insert(t_minutiae, (t_minu_count -1), minutiae[(i -1), :])
min_o_change[(t_minu_count -1), :] = minutiae_o_change
orientations = np.insert(orientations, (t_minu_count -1), temp_o1)
radii = np.insert(radii, (t_minu_count -1), temp_r1)
t_minu_count = t_minu_count + 1
minutiae = t_minutiae.copy()
minu_count = t_minu_count - 1
tmpvec1 = img.shape[0] * np.ones(shape=(minu_count, 1), dtype='float64')
tmpvec2 = np.ones(shape=(minu_count, 1), dtype='float64')
minutiae_for_sc = np.hstack((minutiae[:, 0] / img.shape[1], (tmpvec1 - minutiae[:, 1] + tmpvec2) / img.shape[0]))
dist_m = math.sqrt(dist2(minutiae_for_sc[:, 0:2], minutiae_for_sc[:, 0:2]))
neighbours = np.array([0, 0, 0, 0, 0, 0, 0, 0, 0])
n_i = 1
for i in range(1, (minu_count +1)):
t_a = minutiae[(i -1), 3]
P_x = minutiae[(i -1), 0]
P_y = minutiae[(i -1), 1]
d = np.sort(dist_m[(i -1), :]) # nargout=2
ind = np.argsort(dist_m[(i -1), :])
for j in range(1, (minu_count +1)):
if dist_m[(i -1), (ind[(j -1)] -1)] == 0:
continue
diff_x = minutiae[(i -1), 0] - minutiae[(ind[(j -1)] -1), 0]
diff_y = minutiae[(i -1), 1] - minutiae[(ind[(j -1)] -1), 1]
angle = np.mod(math.atan2(diff_y, diff_x) + np.dot(2, pi), np.dot(2, pi))
theta = np.mod(minutiae[(i -1), 3] - angle + np.dot(2, pi), np.dot(2, pi))
theta_t = np.mod(math.atan2(minutiae[(i -1), 1] - minutiae[(ind[(j -1)] -1), 1], minutiae[(i -1), 0] - minutiae[(ind[(j -1)] -1), 0]), np.dot(2, pi))
ridge_count = 0
p_y = minutiae[(i -1), 1]
p_x = minutiae[(i -1), 0]
t_x = 0
t_y = 0
current = 1
radius = 1
while p_y != minutiae[(ind[(j -1)] -1), 1]:
if thinned[(p_y -1), (p_x -1)] > 0 and current == 0 and (t_x != p_x or t_y != p_y):
current = 1
ridge_count = ridge_count + 1
else:
if thinned[(p_y -1), (p_x -1)] == 0:
current = 0
t_x = p_x
t_y = p_y
p_x = np.around(minutiae[(i -1), 0] - np.dot(radius, math.cos(theta_t)))
p_y = np.around(minutiae[(i -1), 1] - np.dot(radius, math.sin(theta_t)))
radius = radius + 1
co = calc_orient(orientations[(i -1), :], radii[(i -1), :], orientations[(ind[(j -1)] -1), :], radii[(ind[(j -1)] -1), :])
neighbours[(n_i -1), :] = np.array([i, ind[(j -1)], dist_m[(i -1), (ind[(j -1)] -1)], ridge_count, theta_t, minutiae[(ind[(j -1)] -1), 2], minutiae[(ind[(j -1)] -1), 4], co, np.min(abs(minutiae[(i -1), 3] - minutiae[(ind[(j -1)] -1), 3]), np.dot(2, pi) - abs(minutiae[(i -1), 3] - minutiae[(ind[(j -1)] -1), 3]))])
n_i = n_i + 1
pre_singular_min = minutiae
if core_val < 1:
minutiae = np.insert(minutiae, (minu_count + 1 -1), np.array([core_x, core_y, 5, start_t, 0, 1]))
minu_count = minu_count + 1
if dt1 < 1:
minutiae = np.insert(minutiae, (minu_count + 1 -1), np.array([delta1_x, delta1_y, 7, 0, 1, 1]))
minu_count = minu_count + 1
if dt2 < 1:
minutiae = np.insert(minutiae, (minu_count + 1 -1), np.array([delta2_x, delta2_y, 7, 0, 1, 1]))
minu_count = minu_count + 1
if dt3 < 1:
minutiae = np.insert(minutiae, (minu_count + 1 -1), np.array([delta3_x, delta3_y, 7, 0, 1, 1]))
minu_count = minu_count + 1
tmpvec1 = img.shape[0] * np.ones(shape=(minu_count, 1), dtype='float64') #JI: m array of 1's multiplied by image vertical size
tmpvec2 = np.ones(shape=(minu_count, 1), dtype='float64') #JI: m array of 1's
minutiae_for_sc = np.hstack((minutiae[:, 0] / img.shape[1], (tmpvec1 - minutiae[:, 1] + tmpvec2) / img.shape[0])) #convert to x-y cartesian co-ordinates in [0,1]
img_sc = trace_path(thinned, pre_singular_min, 20)
# make minutiae image
if display_flag == 1:
minutiae_img = np.zeros(shape=(img.shape[0], img.shape[1], 3), dtype='uint8')
for i in range(1, (minu_count +1)):
x1 = minutiae[(i -1), 0]
y1 = minutiae[(i -1), 1]
if minutiae[(i -1), 2] == 1:
if minutiae[(i -1), 3] > pi:
for k in range(y1 - 2, (y1 + 2 +1)):
for l in range(x1 - 2, (x1 + 2 +1)):
minutiae_img[(k -1), (l -1), :] = np.array([255, 0, 0]).reshape(1, -1)
else:
for k in range(y1 - 2, (y1 + 2 +1)):
for l in range(x1 - 2, (x1 + 2 +1)):
minutiae_img[(k -1), (l -1), :] = np.array([205, 100, 100]).reshape(1, -1)
elif minutiae[(i -1), 2] == 2:
for k in range(y1 - 2, (y1 + 2 +1)):
for l in range(x1 - 2, (x1 + 2 +1)):
minutiae_img[(k -1), (l -1), :] = np.array([255, 0, 255]).reshape(1, -1)
elif minutiae[(i -1), 2] == 3:
if minutiae[(i -1), 3] > pi:
for k in range(y1 - 2, (y1 + 2 +1)):
for l in range(x1 - 2, (x1 + 2 +1)):
minutiae_img[(k -1), (l -1), :] = np.array([0, 0, 255]).reshape(1, -1)
else:
for k in range(y1 - 2, (y1 + 2 +1)):
for l in range(x1 - 2, (x1 + 2 +1)):
minutiae_img[(k -1), (l -1), :] = np.array([255, 0, 255]).reshape(1, -1)
elif minutiae[(i -1), 2] == 5:
for k in range(y1 - 4, (y1 + 4 +1)):
for l in range(x1 - 4, (x1 + 4 +1)):
minutiae_img[(k -1), (l -1), :] = np.array([0, 255, 0]).reshape(1, -1)
elif minutiae[(i -1), 2] > 5:
for k in range(y1 - 2, (y1 + 2 +1)):
for l in range(x1 - 2, (x1 + 2 +1)):
minutiae_img[(k -1), (l -1), :] = np.array([128, 128, 0]).reshape(1, -1) # gold for delta
# merge thinned image and minutia_img together
combined = minutiae_img.astype('uint8')
for x in range(1, (binim.shape[1] +1)):
for y in range(1, (binim.shape[0] +1)):
if mask[(y -1), (x -1)] == 0:
combined[(y -1), (x -1), :] = np.array([0, 0, 0]).reshape(1, -1)
continue
if (thinned[(y -1), (x -1)]): # binim(y,x))
combined[(y -1), (x -1), :] = np.array([255, 255, 255]).reshape(1, -1)
else:
combined[(y -1), (x -1), :] = np.array([0, 0, 0]).reshape(1, -1)
# end if
if ((minutiae_img[(y -1), (x -1), 2] != 0) or (minutiae_img[(y -1), (x -1), 0] != 0)) or (minutiae_img[(y -1), (x -1), 1] != 0):
combined[(y -1), (x -1), :] = minutiae_img[(y -1), (x -1), :].copy()
if core_val < 1 and YA > 0:
for k in range(YA - 2, (YA + 2 +1)):
for l in range(XA - 2, (XA + 2 +1)):
combined[(k -1), (l -1), :] = np.array([20, 255, 250]).reshape(1, -1)
for k in range(YB - 2, (YB + 2 +1)):
for l in range(XB - 2, (XB + 2 +1)):
combined[(k -1), (l -1), :] = np.array([20, 255, 250]).reshape(1, -1)
plt.figure(1)
#subplot(2,2,1), subimage(img), title(filename)
#subplot(2,3,2), subimage(cimg), title('orientation image 0')
#subplot(2,3,3), subimage(cimg1), title('orientation image 1')
#subplot(2,2,1), subimage(minutiae_img), title('minutiae points')
plt.subplot(1, 2, 1)
plt.subimage(combined)
plt.title(filename)
plt.subplot(1, 2, 2)
plt.subimage(thinned)
plt.title('thinned image')
# subplot(2,2,3), subimage(o_img), title('orientation image (sine)')
# subplot(2,3,5), subimage(eimg), title('f image ')
# subplot(2,2,4), subimage(img_sc), title('orientation image ')
plt.show()
#plotridgeorient(orient_img, 20, img, 2)
print minutiae.shape + '\n'
print minutiae_for_sc.shape + '\n'
print 'minutiae=' + str(minutiae) + '\n'
def dilate(imgname, outfile=None): return erode(imgname, True, outfile)
def extract_finger(filename):
"""Yet to be written"""
blk_sz_c = 24
blk_sz_o = 24
img = mpimg.imread(filename)
#convert to gray scale
if img.ndim == 3: #colour image
img = rgb2gray(img)
yt = 1
yb = img.shape[1]
xr = img.shape[0]
xl = 1
YA = 0
YB = 0
XA = 0
XB = 0
delta_test = 0
for x in range(55):
if np.where(img[x, :] < 200)[0].size < 8:
img[0:x, :] = 255
yt = x + 1
for x in range(224, img.shape[0]):
if np.where(img[x, :] < 200)[0].size < 3:
img[x - 17:img.shape[0], :] = 255
yb = x + 1
break
for y in range(199, img.shape[1]):
if np.where(img[:, y] < 200)[0].size < 1:
img[:, y:img.shape[1]] = 255
xr = y + 1
break
for y in range(75):
if np.where(img[:, y] < 200)[0].size < 1:
img[:, 0:y] = 255
xl = y + 1
enhimg = img.copy()
[cimg1, o1, fimg, bwimg, eimg,
enhimg] = fft_enhance_cubs(enhimg, blk_sz_o / 4)
[cimg, oimg2, fimg, bwimg, eimg,
enhimg] = fft_enhance_cubs(enhimg, blk_sz_o / 2)
[cimg2, oimg, fimg, bwimg, eimg,
enhimg] = fft_enhance_cubs(enhimg, blk_sz_o)
[newim, binim, mask, o_rel,
orient_img] = testfin(enhimg) #testfin is from Dr. Peter Kovesi's code
[cimg, oimg2, fimg, bwimg, eimg, enhimg] = fft_enhance_cubs(img, -1)
[newim, binim, mask, o_rel, orient_img_m] = testfin(enhimg)
theta1 = 0.5
mask_t = mask.copy()
for y in range(19, mask.shape[0] - blk_sz_c * 2 + 1):
for x in range(blk_sz_c, mask.shape[1] - blk_sz_c * 2 + 1):
n_mask = 0
for yy in range(-1, 2):
for xx in range(-1, 2):
y_t = y + yy * blk_sz_c
x_t = x + xx * blk_sz_c
if y_t > 0 and x_t > 0 and (
y_t != y or x_t != x) and mask[y_t - 1,
x_t - 1] == 0:
n_mask = n_mask + 1
if n_mask == 0:
continue
if mask[y - 1, x - 1] == 0 or y > mask.shape[
0] - 20 or y < yt or y > yb or x < xl or x > xr:
cimg2[np.ceil(y / float(blk_sz_c)) - 1,
np.ceil(x / float(blk_sz_c)) - 1] = 255
mask_t[y - 1, x - 1] = 0
continue
for i in range(y, y + 2):
for j in range(x - 9, x + 10):
if i > 0 and j > 0 and i < mask.shape[
0] and j < mask.shape[1] and mask[i - 1,
j - 1] > 0:
continue
else:
cimg2[np.ceil(y / float(blk_sz_c)) - 1,
np.ceil(x / float(blk_sz_c)) - 1] = 255
mask_t[y - 1, x - 1] = 0
break
mask = mask_t.copy()
cimg1[np.where(cimg1 > 0.9)[0]] = 255
cimg1[np.where(cimg > 0.9)[0]] = 255
cimg2[np.where(cimg2 > 0.875)[0]] = 255
#this loop takes 10 minutes
for y in range(1, mask.shape[0] - blk_sz_c * 2 + 1):
for x in range(blk_sz_c, mask.shape[1] - blk_sz_c * 2 + 1):
for i in range(y - 25, y + 26):
for j in range(x - 25, x + 26):
if i > 0 and j > 0 and i < mask.shape[
0] and j < mask.shape[1] and mask[i - 1,
j - 1] > 0:
continue
else:
cimg2[np.ceil(y / float(blk_sz_c)) - 1,
np.ceil(x / float(blk_sz_c)) - 1] = 255
break
for y in range(1, mask.shape[0] - blk_sz_c * 2 + 1):
for x in range(1, mask.shape[1] - blk_sz_c + 1):
if cimg[np.ceil(y / (float(blk_sz_c) / 2)) - 1,
np.ceil(x / (float(blk_sz_c) / 2)) - 1] == 255:
cimg1[np.ceil(y / (float(blk_sz_c) / 4)) - 1,
np.ceil(x / (float(blk_sz_c) / 4)) - 1] = 255
cimg1[np.ceil(y / (float(blk_sz_c) / 4)),
np.ceil(x / (float(blk_sz_c) / 4)) - 1] = 255
cimg1[np.ceil(y / (float(blk_sz_c) / 4)) - 1,
np.ceil(x / (float(blk_sz_c) / 4))] = 255
cimg1[np.ceil(y / (float(blk_sz_c) / 4)),
np.ceil(x / (float(blk_sz_c) / 4))] = 255
for y in range(1, mask.shape[0] - blk_sz_c + 1):
for x in range(1, mask.shape[1] - blk_sz_c + 1):
if cimg2[np.ceil(y / float(blk_sz_c)) - 1,
np.ceil(x / float(blk_sz_c)) - 1] == 255:
if np.ceil(x / (float(blk_sz_c) / 2)) <= cimg.shape[1] - 1:
if np.ceil(y / (float(blk_sz_c) / 2)) <= cimg.shape[0] - 1:
cimg[np.ceil(y / (float(blk_sz_c) / 2)),
np.ceil(x / (float(blk_sz_c) / 2)) - 1] = 255
cimg[np.ceil(y / (float(blk_sz_c) / 2)),
np.ceil(x / (float(blk_sz_c) / 2))] = 255
cimg[np.ceil(y / (float(blk_sz_c) / 2)) - 1,
np.ceil(x / (float(blk_sz_c) / 2))] = 255
else:
xpad = np.ceil(
y / (float(blk_sz_c) / 2)) - (cimg.shape[0] - 1)
else:
ypad = np.ceil(x /
(float(blk_sz_c) / 2)) - (cimg.shape[1] - 1)
cimg = np.hstack((np.vstack((cimg, 255 * np.ones((xpad, cimg.shape[1])))),
np.vstack((255 * np.ones(
(cimg.shape[0], ypad)), 255 * np.ones(
(xpad, ypad))))))
cimg1[np.where(cimg1 < 0.51)[0]] = 255
cimg[np.where(cimg < 0.51)[0]] = 255
cimg2[np.where(cimg2 < 0.51)[0]] = 255
inv_binim = (binim == 0)
pdb.set_trace()
thinned = erode(inv_binim) #What is bwmorph?
mask_t = mask.copy()
if np.where(mask[124:150, 149:250] > 0)[0].size > 0 and np.where(
mask[249:275, 149:250] > 0)[0].size > 0:
mask[149:250, 149:250] = 1
method = -1
core_path = 255
core_y = 0
core_x = 0
core_val = 0
lc = 0
o_img = np.sin(orient_img)
o_img[np.where(mask == 0)[0]] = 1
lower_t = 0.1
v = cimg.min(0)
y = cimg.argmin(0) + 1
dt1 = v.min(0)
x = v.argmin(0) + 1
delta1_y = y[x - 1] * float(blk_sz_c) / 2
delta1_x = x * float(blk_sz_c) / 2
v[x - 1] = 255
v[x] = 255
dt2 = v.min(0)
x = v.argmin(0) + 1
delta2_y = y[x - 1] * float(blk_sz_c) / 2
delta2_x = x * float(blk_sz_c) / 2
v[x - 1] = 255
v[x] = 255
dt3 = v.min(0)
x = v.argmin(0) + 1
delta3_y = y[x - 1] * float(blk_sz_c) / 2
delta3_x = x * float(blk_sz_c) / 2
db = 60
if dt1 < 1 and delta1_y + db < core_y and delta1_y > 15 or dt2 < 1 and delta2_y + db < core_y and delta2_y > 15 or dt3 < 1 and delta3_y + db < core_y \
and delta3_y > 15:
core_val = 255
for y in range(10, o_img.shape[0] - 9):
for x in range(10, o_img.shape[1] - 9):
s1 = 0
t = 10
if y < 50 and x > 250:
t = 11
if y > 38:
yt = 20
else:
yt = 5
if lc > 0.41 and (core_y + 60 < y):
break
if mask[y - 1, x - 1] == 0 or mask[max(y - t,1) - 1, x - 1] == 0 or mask[y - 1, min(x + t, o_img.shape[1]) - 1] == 0 \
or mask[y - 1, max(x - t, 1) - 1] == 0 or mask[max(y - t, 1) - 1, min(x + t, o_img.shape[1]) - 1] == 0 \
and mask[max(y - t, 1) - 1, max(x - t, 1) - 1] == 0 or o_img[y - 1, x - 1] < lc or o_img[y - 1, x - 1] < 0.1:
continue
if dt1 < 1 and delta_y + db < y and delta_y > 15 or dt2 < 1 and delta2_y + db < y and delta2_y > 15 or \
dt3 < 1 and delta3_y + db < y and delta3_y > 15:
continue
test_m = np.min(o_img[0:y - yt,
(np.max((x - 10), 1) -
1):np.min(x + 10, o_img.shape[1])])
if test_m.shape > 0 and np.min(test_m) >= 0.17:
continue
for a in range(y, y + 3):
for b in range(x, x + 2):
s1 = s1 + o_img[a - 1, b - 1]
s1 = float(s1) / 6
s2 = []
i = 1
for a in range(y - 3, y):
for b in range(x, x + 2):
s2[i - 1] = o_img[a - 1, b - 1]
i = i + 1
if min(s2) < lower_t:
s2 = sum(s2) / 6
else:
s2 = s1 #s2 was a list, s1 is a no., this statement makes s2 a number.
s3 = []
i = 1
for a in range(y, y + 3):
for b in range(x + 2, x + 4):
s3[i - 1] = o_img[a - 1, b - 1]
i = i + 1
if min(s3) < lower_t:
s3 = sum(s3) / float(6)
else:
s3 = s1
s4 = []
i = 1
for a in range(y, y + 3):
for b in range(x - 2, x):
s4[i - 1] = o_img[a - 1, b - 1]
i = i + 1
if min(s4) < lower_t:
s4 = sum(s4) / float(6)
else:
s4 = s1
s5 = []
i = 1
for a in range(y - 3, y):
for b in range(x - 2, x):
s5[i - 1] = o_img[a - 1, b - 1]
i = i + 1
if min(s5) < lower_t:
s5 = sum(s5) / float(6)
else:
s5 = s1
s6 = []
for a in range(y - 3, y):
for b in range(x + 2, x + 4):
s6[i - 1] = o_img[a - 1, b - 1]
i = i + 1
if min(s6) < lower_t:
s6 = sum(s6) / float(6)
else:
s6 = s1
if s1 - s2 > core_val:
core_val = s1 - s2
core_x = x
core_y = y
lc = o_img[y - 1, x - 1]
method = 1
if s1 - s3 > core_val:
core_val = s1 - s3
core_x = x
core_y = y
lc = o_img[y - 1, x - 1]
method = 2
if x < 300 and s1 - s4 > core_val:
core_val = s1 - s4
core_x = x
core_y = y
lc = o_img[y - 1, x - 1]
method = 3
if x < 300 and s1 - s5 > core_val:
core_val = s1 - s5
core_x = x
core_y = y
lc = o_img[y - 1, x - 1]
method = 4
if s1 - s6 > core_val:
core_val = s1 - s6
core_x = x
core_y = y
lc = o_img[y - 1, x - 1]
method = 5
#nested for loop ends here
pdb.set_trace() #to be deleted
yt = 0
if core_y > 37:
yt = 20
else:
yt = 5
#Printing is as desired by MATLAB code, reason for printing not known.
print "lc = ", lc
print "core_y = ", core_y
print "core_x = ", core_x
print "method = ", method
test_smooth = 100
if core_y > 0:
test_smooth = np.sum(
np.sum(
o_img[core_y - yt - 6:core_y - yt + 5, core_x - 6:core_x + 5],
0))
if lc > 0.41 and (test_smooth < 109.5 and method != 2
or test_smooth < 100):
start_t = 0
core_val = float(1) / (core_val + 1)
print "core_x = ", core_x
print "core_y = ", core_y
#How to code line 404?
else:
core_x = 0
core_y = 0
core_val = 255
mask = mask_t.copy()
display_flag = 1
ocodes = []
orientations = []
radii = []
sample_intv = 5
path_len = 45
min_o_change = np.zeros([1, floor(float(path_len) / sample_intv)])
#to append more rows in min_o_change, we'll use np.vstack() function
minu_count = 1
minutiae = np.array([0, 0, 0, 0, 0, 1])
# Bibek's code starts here
pdb.set_trace() #to be deleted
# loop through image and find minutiae, ignore certain pixels for border
for y in range(20, (img.shape[0] - 14 + 1)):
for x in range(21, (img.shape[1] - 21 + 1)):
if (thinned[(y - 1), (x - 1)] == 1):
# only continue if pixel is white
# calculate CN from Raymond Thai
CN = 0
sx = 0
sy = 0
for i in range(1, 9):
t1, x1, y1 = p(thinned, x, y, i) # nargout=3
t2, x2, y2 = p(thinned, x, y, i + 1) # nargout=3
CN = CN + abs(t1 - t2)
CN = CN / 2
if ((CN == 1) or (CN == 3)): #&& mask(y,x) > 0
skip = 0
for i in range(y - 5, (y + 5 + 1)):
for j in range(x - 5, (x + 5 + 1)):
if np.logical_and(np.logical_and(i > 0, j > 0),
mask[(i - 1), (j - 1)]) == 0:
skip = 1
if skip == 1:
continue
t_a = np.array([])
c = 0
for e in range(y - 1, (y + 1 + 1)):
for f in range(x - 1, (x + 1 + 1)):
c = c + 1
np.insert(t_a, (c - 1), orient_img_m[(e - 1),
(f - 1)])
m_o = np.median(
t_a
) #orient_img_m(y,x); #oimg(ceil(y/blk_sz_o),ceil(x/blk_sz_o));
m_f = 0 #fimg(ceil(y/blk_sz_o),ceil(x/blk_sz_o));
angle = m_o
if CN == 3:
CN, prog, sx, sy, ang = test_bifurcation(
thinned, x, y, m_o, core_x, core_y) # nargout=5
if prog < 3:
continue
if ang < pi:
m_o = mod(m_o + pi, np.dot(2, pi))
# if ~(ang < pi && min(abs(m_o - ang),2*pi - abs(m_o - ang)) > abs(pi-abs(m_o - ang)))
# m_o = mod(m_o+pi,2*pi);
# end
print 'y = ' + str(y) + '\n'
print 'x = ' + str(x) + '\n'
print 'm_o = ' + str(m_o) + '\n'
print 'ang = ' + str(ang) + '\n'
print 'sy = ' + str(sy) + '\n'
print 'sx = ' + str(sx) + '\n'
print np.min(abs(m_o - ang),
np.dot(2, pi) - abs(m_o - ang))
else:
progress = 0
xx = x
yy = y
pao = -1
pos = 0
while progress < 15 and xx > 1 and yy > 1 and yy < img.shape[
0] and xx < img.shape[1] and pos > -1:
pos = -1
for g in range(1, 9):
ta, xa, ya = p(thinned, xx, yy, g) # nargout=3
tb, xb, yb = p(thinned, xx, yy,
g + 1) # nargout=3
if (ta > tb) and pos == -1 and g != pao:
pos = ta
if g < 5:
pao = 4 + g
else:
pao = mod(4 + g, 9) + 1
xx = xa
yy = ya
progress = progress + 1
if progress < 10:
continue
if mod(atan2(y - yy, xx - x), np.dot(2, pi)) > pi:
m_o = m_o + pi
minutiae[(minu_count - 1), :] = np.array(
[x, y, CN, m_o, m_f, 1]).reshape(1, -1)
min_path_index[(minu_count - 1), :] = np.array([sx, sy])
minu_count = minu_count + 1
#end of if loop if pixel white
# enf of for y
# end of for x
# Identify the ellipse enclosed by the largest rectangle spanned by the
# set of initially detected minutiaes (4/8/2006 - Paul Kwan)
min_x = np.min(minutiae[:, 0])
max_x = np.max(minutiae[:, 0])
min_y = np.min(minutiae[:, 1]) / 1.1
max_y = np.dot(np.max(minutiae[:, 1]), 1.1)
mid_x = (min_x + max_x) / 2
mid_y = (min_y + max_y) / 2
if (max_x - min_x) > (max_y - min_y): # major axis is on the x-axis
major_axis_len = max_x - min_x # Refer to figure
M_Y1 = (max_y - min_y) / 2
F1_Y1 = major_axis_len / 2
M_F1 = sqrt(F1_Y1**2 - M_Y1**2) # Use Pythagoras' theorem
F1_x = mid_x - M_F1
F1_y = mid_y
F2_x = mid_x + M_F1
F2_y = mid_y
else:
major_axis_len = max_y - min_y # Refer to figure
M_X1 = (max_x - min_x) / 2
F1_X1 = major_axis_len / 2
M_F1 = sqrt(F1_X1**2 - M_X1**2) # Use Pythagoras' theorem
F1_x = mid_x
F1_y = mid_y + M_F1
F2_x = mid_x
F2_y = mid_y - M_F1
clean_list = np.array([])
c_index = 1
thinned_old = thinned.copy()
minu_count = minu_count - 1
# Filter potential false endings by excluding those outside of the
# elliptical region
t_minutiae = np.array([])
t_minu_count = 1
t_mpi = np.array([])
pre_minu_count1 = minu_count
for i in range(1, (minu_count + 1)):
X = minutiae[(i - 1), 0]
Y = minutiae[(i - 1), 1]
rc = 0
for y in range(max(Y - 2, 1), (min(Y + 2, binim.shape[0]) + 1)):
if rc > 0:
break
for x in range(max(X - 2, 1), (min(X + 2, binim.shape[1]) + 1)):
if mask[(y - 1), (x - 1)] == 0:
rc = rc + 1
break
if rc > 0:
continue
else:
np.insert(t_minutiae, (t_minu_count - 1), minutiae[(i - 1), :])
np.insert(t_mpi, (t_minu_count - 1), min_path_index[(i - 1), :])
t_minu_count = t_minu_count + 1
minutiae = t_minutiae.copy()
min_path_index = t_mpi.copy()
minu_count = minutiae.shape[0]
t_minu_count = 1
t_minutiae = np.array([])
pre_minu_count2 = minu_count
print 'pre_minu_count2 =' + str(pre_minu_count2) + '\n'
dist_m = dist2(minutiae[:, 0:2], minutiae[:, 0:2])
#dist_m(find(dist_m == 0)) == 1000;
dist_test = 49
#if minu_count > 50
# dist_test=81;
#end
for i in range(1, (minu_count + 1)):
reject_flag = 0
t_a = minutiae[(i - 1), 3]
P_x = minutiae[(i - 1), 0]
P_y = minutiae[(i - 1), 1]
for j in range(i + 1, (minu_count + 1)):
if dist_m[(i - 1), (j - 1)] <= dist_test:
reject_flag = 1
if reject_flag == 0 and mask[(P_y - 1), (P_x - 1)] > 0:
reverse_p = 0
if min_path_index[(i - 1), 0] == 0:
x = P_x
y = P_y
else:
x = min_path_index[(i - 1), 0]
y = min_path_index[(i - 1), 1]
iter_ = 0
p1x = P_x
p1y = P_y
p2x = P_x
p2y = P_y
minutiae_o_change = np.array([])
x1 = x
y1 = y
xs = x
ys = y
p_uv = np.array([(x / math.sqrt(x**2 + y**2)),
(y / math.sqrt(x**2 + y**2))])
iter_ = 0
for m in range(1, (path_len + 1)):
iter_ = iter_ + 1
cn = 0
for ii in range(1, 9):
t1, x_A, y_A = p(thinned, x1, y1, ii) # nargout=3
t2, x_B, y_B = p(thinned, x1, y1, ii + 1) # nargout=3
cn = cn + abs(t1 - t2)
cn = cn / 2
if (cn != 3 and cn != 4) or m == 1: #&& mask(x1,y1) > 0
for n in range(1, 9):
if reverse_p == 0 or iter_ > 1:
ta, xa, ya = p(thinned, x1, y1, n) # nargout=3
else:
ta, xa, ya = p(thinned, x1, y1, 9 - n) # nargout=3
if ta == 1 and (xa != p1x
or ya != p1y) and (xa != x or ya != y):
p1x = x1
p1y = y1
x1 = xa
y1 = ya
break
if np.mod(iter_, sample_intv) == 0:
if xs == x1 and ys == y1:
minutiae_o_change = np.insert(
minutiae_o_change, (iter_ / sample_intv - 1), -1)
else:
norm_s = math.sqrt((x1 - xs)**2 + (y1 - ys)**2)
xv = (x1 - xs) / norm_s
yv = (y1 - ys) / norm_s
ptx = x - xs
pty = y - ys
norm_p = math.sqrt((ptx)**2 + (pty)**2)
ptx = ptx / norm_p
pty = pty / norm_p
#inner product
i_prod = (np.dot(xv, ptx) + np.dot(yv, pty))
minutiae_o_change = np.insert(
minutiae_o_change, (iter_ / sample_intv - 1),
math.acos(i_prod))
xs = x1
ys = y1
minutiae_o_change = np.insert(minutiae_o_change, 0, 1)
# [temp_o1, temp_r1, mx1, my1] = tico(binim, P_x, P_y, oimg, orient_img_m, o_rel, [12 24 36 48 60 72 84] , [14 18 24 28 30 34 28] , 1, minutiae(i,4), blk_sz_o);
temp_o1, temp_r1, mx1, my1 = tico(
binim, P_x, P_y, oimg, orient_img_m, o_rel,
np.array([12, 24, 36, 48, 60, 72, 84]),
np.array([14, 18, 24, 28, 30, 34,
28]), 0, -1, blk_sz_o) # nargout=4
t_minutiae = np.insert(t_minutiae, (t_minu_count - 1),
minutiae[(i - 1), :])
min_o_change[(t_minu_count - 1), :] = minutiae_o_change
orientations = np.insert(orientations, (t_minu_count - 1), temp_o1)
radii = np.insert(radii, (t_minu_count - 1), temp_r1)
t_minu_count = t_minu_count + 1
minutiae = t_minutiae.copy()
minu_count = t_minu_count - 1
tmpvec1 = img.shape[0] * np.ones(shape=(minu_count, 1), dtype='float64')
tmpvec2 = np.ones(shape=(minu_count, 1), dtype='float64')
minutiae_for_sc = np.hstack(
(minutiae[:, 0] / img.shape[1],
(tmpvec1 - minutiae[:, 1] + tmpvec2) / img.shape[0]))
dist_m = math.sqrt(dist2(minutiae_for_sc[:, 0:2], minutiae_for_sc[:, 0:2]))
neighbours = np.array([0, 0, 0, 0, 0, 0, 0, 0, 0])
n_i = 1
for i in range(1, (minu_count + 1)):
t_a = minutiae[(i - 1), 3]
P_x = minutiae[(i - 1), 0]
P_y = minutiae[(i - 1), 1]
d = np.sort(dist_m[(i - 1), :]) # nargout=2
ind = np.argsort(dist_m[(i - 1), :])
for j in range(1, (minu_count + 1)):
if dist_m[(i - 1), (ind[(j - 1)] - 1)] == 0:
continue
diff_x = minutiae[(i - 1), 0] - minutiae[(ind[(j - 1)] - 1), 0]
diff_y = minutiae[(i - 1), 1] - minutiae[(ind[(j - 1)] - 1), 1]
angle = np.mod(
math.atan2(diff_y, diff_x) + np.dot(2, pi), np.dot(2, pi))
theta = np.mod(minutiae[(i - 1), 3] - angle + np.dot(2, pi),
np.dot(2, pi))
theta_t = np.mod(
math.atan2(
minutiae[(i - 1), 1] - minutiae[(ind[(j - 1)] - 1), 1],
minutiae[(i - 1), 0] - minutiae[(ind[(j - 1)] - 1), 0]),
np.dot(2, pi))
ridge_count = 0
p_y = minutiae[(i - 1), 1]
p_x = minutiae[(i - 1), 0]
t_x = 0
t_y = 0
current = 1
radius = 1
while p_y != minutiae[(ind[(j - 1)] - 1), 1]:
if thinned[(p_y - 1),
(p_x - 1)] > 0 and current == 0 and (t_x != p_x
or t_y != p_y):
current = 1
ridge_count = ridge_count + 1
else:
if thinned[(p_y - 1), (p_x - 1)] == 0:
current = 0
t_x = p_x
t_y = p_y
p_x = np.around(minutiae[(i - 1), 0] -
np.dot(radius, math.cos(theta_t)))
p_y = np.around(minutiae[(i - 1), 1] -
np.dot(radius, math.sin(theta_t)))
radius = radius + 1
co = calc_orient(orientations[(i - 1), :], radii[(i - 1), :],
orientations[(ind[(j - 1)] - 1), :],
radii[(ind[(j - 1)] - 1), :])
neighbours[(n_i - 1), :] = np.array([
i, ind[(j - 1)], dist_m[(i - 1), (ind[(j - 1)] - 1)],
ridge_count, theta_t, minutiae[(ind[(j - 1)] - 1), 2],
minutiae[(ind[(j - 1)] - 1), 4], co,
np.min(
abs(minutiae[(i - 1), 3] -
minutiae[(ind[(j - 1)] - 1), 3]),
np.dot(2, pi) - abs(minutiae[(i - 1), 3] -
minutiae[(ind[(j - 1)] - 1), 3]))
])
n_i = n_i + 1
pre_singular_min = minutiae
if core_val < 1:
minutiae = np.insert(minutiae, (minu_count + 1 - 1),
np.array([core_x, core_y, 5, start_t, 0, 1]))
minu_count = minu_count + 1
if dt1 < 1:
minutiae = np.insert(minutiae, (minu_count + 1 - 1),
np.array([delta1_x, delta1_y, 7, 0, 1, 1]))
minu_count = minu_count + 1
if dt2 < 1:
minutiae = np.insert(minutiae, (minu_count + 1 - 1),
np.array([delta2_x, delta2_y, 7, 0, 1, 1]))
minu_count = minu_count + 1
if dt3 < 1:
minutiae = np.insert(minutiae, (minu_count + 1 - 1),
np.array([delta3_x, delta3_y, 7, 0, 1, 1]))
minu_count = minu_count + 1
tmpvec1 = img.shape[0] * np.ones(
shape=(minu_count, 1),
dtype='float64') #JI: m array of 1's multiplied by image vertical size
tmpvec2 = np.ones(shape=(minu_count, 1),
dtype='float64') #JI: m array of 1's
minutiae_for_sc = np.hstack(
(minutiae[:, 0] / img.shape[1], (tmpvec1 - minutiae[:, 1] + tmpvec2) /
img.shape[0])) #convert to x-y cartesian co-ordinates in [0,1]
img_sc = trace_path(thinned, pre_singular_min, 20)
# make minutiae image
if display_flag == 1:
minutiae_img = np.zeros(shape=(img.shape[0], img.shape[1], 3),
dtype='uint8')
for i in range(1, (minu_count + 1)):
x1 = minutiae[(i - 1), 0]
y1 = minutiae[(i - 1), 1]
if minutiae[(i - 1), 2] == 1:
if minutiae[(i - 1), 3] > pi:
for k in range(y1 - 2, (y1 + 2 + 1)):
for l in range(x1 - 2, (x1 + 2 + 1)):
minutiae_img[(k - 1), (l - 1), :] = np.array(
[255, 0, 0]).reshape(1, -1)
else:
for k in range(y1 - 2, (y1 + 2 + 1)):
for l in range(x1 - 2, (x1 + 2 + 1)):
minutiae_img[(k - 1), (l - 1), :] = np.array(
[205, 100, 100]).reshape(1, -1)
elif minutiae[(i - 1), 2] == 2:
for k in range(y1 - 2, (y1 + 2 + 1)):
for l in range(x1 - 2, (x1 + 2 + 1)):
minutiae_img[(k - 1),
(l - 1), :] = np.array([255, 0, 255
]).reshape(1, -1)
elif minutiae[(i - 1), 2] == 3:
if minutiae[(i - 1), 3] > pi:
for k in range(y1 - 2, (y1 + 2 + 1)):
for l in range(x1 - 2, (x1 + 2 + 1)):
minutiae_img[(k - 1), (l - 1), :] = np.array(
[0, 0, 255]).reshape(1, -1)
else:
for k in range(y1 - 2, (y1 + 2 + 1)):
for l in range(x1 - 2, (x1 + 2 + 1)):
minutiae_img[(k - 1), (l - 1), :] = np.array(
[255, 0, 255]).reshape(1, -1)
elif minutiae[(i - 1), 2] == 5:
for k in range(y1 - 4, (y1 + 4 + 1)):
for l in range(x1 - 4, (x1 + 4 + 1)):
minutiae_img[(k - 1),
(l - 1), :] = np.array([0, 255, 0
]).reshape(1, -1)
elif minutiae[(i - 1), 2] > 5:
for k in range(y1 - 2, (y1 + 2 + 1)):
for l in range(x1 - 2, (x1 + 2 + 1)):
minutiae_img[(k - 1), (l - 1), :] = np.array(
[128, 128, 0]).reshape(1, -1) # gold for delta
# merge thinned image and minutia_img together
combined = minutiae_img.astype('uint8')
for x in range(1, (binim.shape[1] + 1)):
for y in range(1, (binim.shape[0] + 1)):
if mask[(y - 1), (x - 1)] == 0:
combined[(y - 1),
(x - 1), :] = np.array([0, 0, 0]).reshape(1, -1)
continue
if (thinned[(y - 1), (x - 1)]): # binim(y,x))
combined[(y - 1),
(x - 1), :] = np.array([255, 255,
255]).reshape(1, -1)
else:
combined[(y - 1),
(x - 1), :] = np.array([0, 0, 0]).reshape(1, -1)
# end if
if ((minutiae_img[(y - 1), (x - 1), 2] != 0) or
(minutiae_img[(y - 1), (x - 1), 0] != 0)) or (
minutiae_img[(y - 1), (x - 1), 1] != 0):
combined[(y - 1),
(x - 1), :] = minutiae_img[(y - 1),
(x - 1), :].copy()
if core_val < 1 and YA > 0:
for k in range(YA - 2, (YA + 2 + 1)):
for l in range(XA - 2, (XA + 2 + 1)):
combined[(k - 1),
(l - 1), :] = np.array([20, 255,
250]).reshape(1, -1)
for k in range(YB - 2, (YB + 2 + 1)):
for l in range(XB - 2, (XB + 2 + 1)):
combined[(k - 1),
(l - 1), :] = np.array([20, 255,
250]).reshape(1, -1)
plt.figure(1)
#subplot(2,2,1), subimage(img), title(filename)
#subplot(2,3,2), subimage(cimg), title('orientation image 0')
#subplot(2,3,3), subimage(cimg1), title('orientation image 1')
#subplot(2,2,1), subimage(minutiae_img), title('minutiae points')
plt.subplot(1, 2, 1)
plt.subimage(combined)
plt.title(filename)
plt.subplot(1, 2, 2)
plt.subimage(thinned)
plt.title('thinned image')
# subplot(2,2,3), subimage(o_img), title('orientation image (sine)')
# subplot(2,3,5), subimage(eimg), title('f image ')
# subplot(2,2,4), subimage(img_sc), title('orientation image ')
plt.show()
#plotridgeorient(orient_img, 20, img, 2)
print minutiae.shape + '\n'
print minutiae_for_sc.shape + '\n'
print 'minutiae=' + str(minutiae) + '\n'