def horizontal_pivot(img, table_file, plot, config):
copy_img = img.copy()
dimension = config.getint('horizontalpivot', 'eroseDimension')
bandwidth = config.getfloat('horizontalpivot', 'bandwidth')
erosion = enhance(copy_img, table_file, 1, dimension, None, None, config)
# gray = cv2.cvtColor(norm_img, cv2.COLOR_BGR2GRAY)
hei, wid = erosion.shape
#print('shape', hei, wid)
_, norm_img = cv2.threshold(erosion, 0, 1,
cv2.THRESH_BINARY | cv2.THRESH_OTSU)
'''projecting'''
hist = np.sum(norm_img, axis=0, dtype=np.int32)
invert_hist = hei - hist
# logging.info('hist')
list_bin = np.arange(0, wid)
calculus_hist = np.repeat(list_bin, invert_hist)
'''extremas of probability density'''
is_blank = False
try:
mi, ma, s, prb = kernel_density_estimate(calculus_hist, wid, bandwidth)
except:
is_blank = True
return None, erosion, is_blank
#print('im her++++++++++++++++++++++++++++++++++++++++')
estimate_wid = np.std(ma)
#print('estimate wid ', estimate_wid)
list_arg_pivot_min = mi
# logging.info('list arg pivot min')
# logging.info(len(list_arg_pivot_min))
# check_cross = config.getint('horizontalpivot', 'checkcross')
# if check_cross:config.read
# gray = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)
# _, thesh = cv2.threshold(gray, 0, 1, cv2.THRESH_BINARY | cv2.THRESH_OTSU)
# list_arg_pivot_min = checkCrossing(thesh, mi)
''' smooth pivot'''
# logging.info('check smoothed hist')
smoothed_bin = check_smooth_hist(copy_img, list_arg_pivot_min, ma,
invert_hist, hei, config)
# logging.info(len(smoothed_bin))
# logging.info(smoothed_bin)
nearest_neighbor = config.getint('horizontalpivot', 'nearestNeighbor')
reduce_arg_min = reduction(copy_img, invert_hist, smoothed_bin, mi, ma,
nearest_neighbor, estimate_wid, config)
# logging.info(reduce_arg_min)
check_cross = False #config.getint('horizontalpivot', 'checkcross')
if check_cross:
gray = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)
_, thesh = cv2.threshold(gray, 0, 1,
cv2.THRESH_BINARY | cv2.THRESH_OTSU)
final_pivot = checkCrossing(thesh, reduce_arg_min)
else:
final_pivot = reduce_arg_min
# logging.info(len(reduce_arg_min))
# logging.info('list reduced arg min')
# logging.info(final_pivot)
'''plot'''
if (plot):
fig = plt.figure()
ax1 = plt.subplot2grid((2, 3), (0, 0))
ax1.set_xlim([0, wid])
plt.hist(calculus_hist, bins=50, normed=1)
plt.legend()
plt.plot(s, prb)
ax2 = plt.subplot2grid((2, 3), (0, 1), colspan=2)
ax2.set_xlim([0, wid])
# prb_unnorm = prb*hei
plt.plot(s, prb, 'b', s[ma], prb[ma], 'go', s[mi], prb[mi], 'ro')
# for i in final_pivot:
# cv2.line(norm_img, (i, hei), (i, 0), (255, 0, 0), 1)
ax3 = plt.subplot2grid((2, 3), (1, 0))
ax3.set_xlim([0, wid])
plt.imshow(erosion)
# ax4 = plt.subplot2grid((2, 3), (1, 0))
# ax4.set_xlim([0, wid])
# ax4.set_title('smoothed bin')
# plt.plot(list_bin, invert_hist, 'b',
# list_bin[smoothed_bin], hist[smoothed_bin], 'ro')
for i in final_pivot:
cv2.line(copy_img, (i, hei), (i, 0), (255, 0, 0), 2)
ax5 = plt.subplot2grid((2, 3), (1, 1), colspan=2)
ax5.set_xlim([0, wid])
plt.imshow(copy_img)
figManager = plt.get_current_fig_manager()
figManager.resize(*figManager.window.maxsize())
plt.tight_layout()
plt.show()
# fig.savefig('{}.PNG'.format(table_file[:-4]))
return final_pivot, erosion, is_blank
def horizontal_pivot(img, table_file, plot):
# img = cv2.imread(table_file)
# img1 = np.asarray(img.copy())
erosion = enhance(img, table_file, 0)
# gray = cv2.cvtColor(norm_img, cv2.COLOR_BGR2GRAY)
hei, wid = erosion.shape
print('shape', hei, wid)
_, norm_img = cv2.threshold(erosion, 0, 1,
cv2.THRESH_BINARY | cv2.THRESH_OTSU)
'''projecting'''
hist = np.sum(norm_img, axis=0, dtype=np.int32)
invert_hist = hei - hist
print('hi', wid)
logging.info('hist')
logging.info(hist)
list_bin = np.arange(0, wid)
num_ones = np.count_nonzero(norm_img)
calculus_hist = np.repeat(list_bin, invert_hist)
'''extremas of probability density'''
mi, ma, s, prb = kernel_density_estimate(calculus_hist, wid, 20)
thresh_interval_right = 160
thresh_interval_left = 120
list_arg_pivot_min = mi[(mi < wid - thresh_interval_right)
& (mi > thresh_interval_left)]
logging.info('list arg pivot min')
logging.info(list_arg_pivot_min.shape)
nearest_neighbor = 120
'''threshold for get low value pivot'''
max_invert_hist = max(invert_hist[ma])
logging.info('max_invert_hist')
logging.info(max_invert_hist)
'''filter low hist value '''
list_arg_pivot_min = list_arg_pivot_min[
invert_hist[list_arg_pivot_min] < max_invert_hist * 2 / 3]
logging.info('list arg pivot min')
logging.info(list_arg_pivot_min)
reduce_arg_min = reduction(list_arg_pivot_min, nearest_neighbor)
print(len(reduce_arg_min))
logging.info('list reduced arg min')
logging.info(reduce_arg_min)
''' smooth pivot'''
smoothed_bin = check_smooth_hist(reduce_arg_min, ma, invert_hist)
logging.info('check smoothed hist')
logging.info(smoothed_bin)
'''plot'''
if (plot):
fig = plt.figure()
ax1 = plt.subplot(231)
ax1.set_xlim([0, wid])
plt.hist(calculus_hist, bins=50, normed=1)
plt.legend()
plt.plot(s, prb)
ax2 = plt.subplot(232)
ax2.set_xlim([0, wid])
plt.plot(s, prb, 'b', s[ma], prb[ma], 'go', s[mi], prb[mi], 'ro')
for i in smoothed_bin:
cv2.line(norm_img, (i, hei), (i, 0), (255, 0, 0), 2)
ax3 = plt.subplot(233)
ax3.set_xlim([0, wid])
plt.imshow(erosion)
ax4 = plt.subplot(234)
ax4.set_xlim([0, wid])
plt.plot(list_bin, invert_hist, 'b', list_bin[smoothed_bin],
hist[smoothed_bin], 'ro')
for i in smoothed_bin:
cv2.line(img, (i, hei), (i, 0), (255, 0, 0), 4)
ax5 = plt.subplot(235)
ax5.set_xlim([0, wid])
plt.imshow(img)
figManager = plt.get_current_fig_manager()
figManager.window.showMaximized()
plt.tight_layout()
plt.show()
return smoothed_bin
def horizontal_pivot(img, table_file, plot):
dimension = 25
erosion = enhance(img, table_file, 1, dimension)
# gray = cv2.cvtColor(norm_img, cv2.COLOR_BGR2GRAY)
hei, wid = erosion.shape
#print('shape', hei, wid)
_, norm_img = cv2.threshold(erosion, 0, 1,
cv2.THRESH_BINARY | cv2.THRESH_OTSU)
'''projecting'''
hist = np.sum(norm_img, axis=0, dtype=np.int32)
invert_hist = hei - hist
#print('hi', wid)
logging.info('hist')
list_bin = np.arange(0, wid)
calculus_hist = np.repeat(list_bin, invert_hist)
'''extremas of probability density'''
is_blank = False
try:
mi, ma, s, prb = kernel_density_estimate(calculus_hist, wid, 1.5)
except:
is_blank = True
return None, erosion, is_blank
estimate_wid = np.std(ma)
#print('estimate wid ', estimate_wid)
list_arg_pivot_min = mi
logging.info('list arg pivot min')
logging.info(len(list_arg_pivot_min))
''' smooth pivot'''
logging.info('check smoothed hist')
smoothed_bin = check_smooth_hist(list_arg_pivot_min, ma, invert_hist, hei)
logging.info(len(smoothed_bin))
logging.info(smoothed_bin)
nearest_neighbor = 15
reduce_arg_min = reduction(invert_hist, smoothed_bin, mi, ma,
nearest_neighbor, estimate_wid)
# reduce_arg_min = smoothed_bin
logging.info(len(reduce_arg_min))
logging.info('list reduced arg min')
logging.info(reduce_arg_min)
'''plot'''
if (plot):
fig = plt.figure()
ax1 = plt.subplot2grid((2, 3), (0, 0))
ax1.set_xlim([0, wid])
plt.hist(calculus_hist, bins=50, normed=1)
plt.legend()
plt.plot(s, prb)
ax2 = plt.subplot2grid((2, 3), (0, 1), colspan=2)
ax2.set_xlim([0, wid])
plt.plot(s, prb, 'b', s[ma], prb[ma], 'go', s[mi], prb[mi], 'ro')
for i in reduce_arg_min:
cv2.line(norm_img, (i, hei), (i, 0), (255, 0, 0), 1)
ax3 = plt.subplot2grid((2, 3), (1, 0))
ax3.set_xlim([0, wid])
plt.imshow(erosion)
for i in reduce_arg_min:
cv2.line(img, (i, hei), (i, 0), (255, 0, 0), 1)
ax5 = plt.subplot2grid((2, 3), (1, 1), colspan=2)
ax5.set_xlim([0, wid])
plt.imshow(img)
figManager = plt.get_current_fig_manager()
figManager.window.showMaximized()
plt.tight_layout()
plt.show()
# fig.savefig('{}.PNG'.format(table_file[:-4]))
return reduce_arg_min, erosion, is_blank
def vertical_pivot(img, table_file, plot):
erosion = enhance(img, table_file, 1)
hei, wid = erosion.shape
print('shape', hei, wid)
_, norm_img = cv2.threshold(erosion, 0, 1,
cv2.THRESH_BINARY | cv2.THRESH_OTSU)
'''projecting'''
hist = np.sum(norm_img, axis=1, dtype=np.int32)
invert_hist = wid - hist
logging.info('hei', hei)
list_bin = np.arange(0, hei)
num_ones = np.count_nonzero(norm_img)
calculus_hist = np.repeat(list_bin, invert_hist)
'''extremas of probability density'''
mi, ma, s, prb = kernel_density_estimate(calculus_hist, hei, 4)
thresh_interval = 50
list_arg_pivot_min = mi[mi > thresh_interval]
nearest_neighbor = 20
'''threshold for get low value pivot'''
max_invert_hist = max(invert_hist[ma])
print('max_invert_hist', max_invert_hist)
list_arg_pivot_min = list_arg_pivot_min[
invert_hist[list_arg_pivot_min] < max_invert_hist / 2]
reduce_arg_min = reduction(list_arg_pivot_min, nearest_neighbor)
'''sorting pivot'''
'''adjust highest pivot'''
if len(reduce_arg_min) >= 1:
reduce_arg_min[0] -= 10
print(len(reduce_arg_min))
logging.info('list reduced arg min')
logging.info(list_arg_pivot_min)
'''plot'''
if (plot):
'''histogram'''
fig = plt.figure()
ax1 = plt.subplot(231)
# ax1.set_xlim([0, wid])
ax1.set_ylim([hei, 0])
plt.hist(calculus_hist, bins=50, normed=1, orientation="horizontal")
plt.legend()
plt.plot(prb, s)
'''kde'''
ax2 = plt.subplot(232)
# ax2.set_xlim([0, wid])
ax2.set_ylim([hei, 0])
plt.plot(prb, s, 'b', prb[ma], s[ma], 'go', prb[mi], s[mi], 'ro')
'''erosion image'''
for i in reduce_arg_min:
cv2.line(norm_img, (wid, i), (0, i), (255, 0, 0), 2)
ax3 = plt.subplot(233)
ax3.set_xlim([0, wid])
ax3.set_ylim([hei, 0])
plt.imshow(erosion)
''''''
# ax4 = plt.subplot(234)
# ax4.set_xlim([0, wid])
# plt.plot(list_bin, invert_hist, 'b',
# # list_bin[mi], hist[mi], 'ro',
# list_bin[reduce_arg_min], hist[reduce_arg_min], 'ro')
'''image'''
for i in reduce_arg_min:
cv2.line(img, (wid, i), (0, i), (255, 0, 0), 4)
ax5 = plt.subplot(235)
ax5.set_xlim([0, wid])
ax5.set_ylim([hei, 0])
plt.imshow(img)
figManager = plt.get_current_fig_manager()
figManager.window.showMaximized()
plt.tight_layout()
plt.show()
return reduce_arg_min