def my_application():
from gamera.toolkits.ocr.classes import Page
# Make the image variable a global variable
# IF YOU DON'T DO THIS, THE WINDOW WILL DISAPPEAR!
global image, ots, ots_up, ots_mid, box_image, ots_low
print sys.argv[-1]
#image_color = load_image(sys.argv[-1])
picture = mh.imread(sys.argv[-1], as_grey=True)
picture = picture.astype(np.uint8)
# Load the image
image = numpy_io.from_numpy(picture)#load_image(sys.argv[-1]
#image = image_color.to_greyscale()
# Display the image in a window
#image.simple_sharpen()
ots = image.otsu_threshold()
thresh = image.otsu_find_threshold()
#thresh_low = int(thresh* 1.05)
thresh_mid = thresh_plus = int(thresh * 1.0)
ots_mid = image.threshold(thresh_mid)
my_filter(ots_mid)
ots_up = ots_mid.image_copy()#image.threshold(thresh_plus)
#ots_up.remove_border()
# ots_low = image.threshold(thresh_low)
print "OTSU: " + str(thresh)
print "OTSU Mid: " + str(thresh_mid)
print "OTSU * 1.15: " + str(thresh_plus)
#ots_low = filter_and_bbox(ots_low)
#ots = filter_and_bbox(ots)
(ots_up,ots_mid) = filter_and_bbox(ots_up)
#ots_mid = only_app_crit(ots_mid,x)
#print "x is " + str(x)
ots_up.display("OTSU-border")
ots_mid.display()
def my_application():
from gamera.toolkits.ocr.classes import Page
# Make the image variable a global variable
# IF YOU DON'T DO THIS, THE WINDOW WILL DISAPPEAR!
global image, ots, ots_up, ots_mid, box_image, ots_low
print sys.argv[-1]
#image_color = load_image(sys.argv[-1])
picture = mh.imread(sys.argv[-1], as_grey=True)
picture = picture.astype(np.uint8)
# Load the image
image = numpy_io.from_numpy(picture) #load_image(sys.argv[-1]
#image = image_color.to_greyscale()
# Display the image in a window
#image.simple_sharpen()
ots = image.otsu_threshold()
thresh = image.otsu_find_threshold()
thresh_low = int(thresh * 0.90)
thresh_mid = int(thresh * 1.05)
thresh_plus = int(thresh * 1.18)
ots_up = image.threshold(thresh_plus)
ots_mid = image.threshold(thresh_mid)
ots_low = image.threshold(thresh_low)
print "OTSU: " + str(thresh)
print "OTSU Mid: " + str(thresh_mid)
print "OTSU * 1.15: " + str(thresh_plus)
ots_low = filter_and_bbox(ots_low)
ots = filter_and_bbox(ots)
ots_up = filter_and_bbox(ots_up)
ots_up.display("OTSU+")
#ots_mid = show_lines(p)
#box_image.display()
#my_filter(ots_mid)
ots_up.display("OTSU+FIL")
ots_mid = filter_and_bbox(ots_mid)
ots_mid.display("MID")
ots.display("OTSU")
ots_low.display("OTSU_LOW")
image.display()
def my_application():
from gamera.toolkits.ocr.classes import Page
# Make the image variable a global variable
# IF YOU DON'T DO THIS, THE WINDOW WILL DISAPPEAR!
global image, ots, ots_up, ots_mid, box_image, ots_low
print sys.argv[-1]
#image_color = load_image(sys.argv[-1])
picture = mh.imread(sys.argv[-1], as_grey=True)
picture = picture.astype(np.uint8)
# Load the image
image = numpy_io.from_numpy(picture)#load_image(sys.argv[-1]
#image = image_color.to_greyscale()
# Display the image in a window
#image.simple_sharpen()
ots = image.otsu_threshold()
thresh = image.otsu_find_threshold()
thresh_low = int(thresh* 0.90)
thresh_mid = int(thresh * 1.05)
thresh_plus = int(thresh * 1.18)
ots_up = image.threshold(thresh_plus)
ots_mid = image.threshold(thresh_mid)
ots_low = image.threshold(thresh_low)
print "OTSU: " + str(thresh)
print "OTSU Mid: " + str(thresh_mid)
print "OTSU * 1.15: " + str(thresh_plus)
ots_low = filter_and_bbox(ots_low)
ots = filter_and_bbox(ots)
ots_up = filter_and_bbox(ots_up)
ots_up.display("OTSU+")
#ots_mid = show_lines(p)
#box_image.display()
#my_filter(ots_mid)
ots_up.display("OTSU+FIL")
ots_mid = filter_and_bbox(ots_mid)
ots_mid.display("MID")
ots.display("OTSU")
ots_low.display("OTSU_LOW")
image.display()
def performGreekOCR(options):
import mahotas as mh
import numpy as np
from pylab import imshow, gray, show
#from os import path
from gamera.plugins import numpy_io
# features = ["aspect_ratio", "volume64regions", "moments", "nholes_extended"]
#I think these are size-invariant
# features = ["aspect_ratio","moments","nholes","nholes_extended","skeleton_features","top_bottom","volume","volume16regions","volume64regions","zernike_moments"]
MAX_CCS = 6500
features = ["aspect_ratio","moments","ncols_feature","nholes","nholes_extended","nrows_feature","skeleton_features","top_bottom","volume","volume16regions","volume64regions","zernike_moments"]
image_files = []
g = GreekOCR(splits=options["split"],feats=features)
g.mode = options["mode"] + "body"
g.autogroup = options["autogroup"]
g.debug = options["debug"]
g.load_trainingdata(options['trainingdata'])
g_appcrit = GreekOCR(splits=options["split"], feats=features)
g_appcrit.mode = options["mode"] + "appcrit"
g_appcrit.autogroup = options["autogroup"]
g_appcrit.debug = options["debug"]
g_appcrit.load_trainingdata(options['trainingdata'])
if options["hocrfile"]:
g.hocr = (options["hocrfile"])
if options["settingsfile"]:
g.load_settings(options["settingsfile"])
g_appcrit.load_settings(options["settingsfile"])
if options["otsu"]:
otsu_factors_string = options["otsu"].split(',')
otsu_factors = [float(x) for x in otsu_factors_string]
else:
otsu_factors = [0]
if options["directory"]:
image_files = os.listdir(options["directory"])
image_files = [os.path.join(options["directory"],x) for x in image_files]
test = re.compile(".png$",re.IGNORECASE)
image_files = filter(test.search, image_files)
image_files.sort()
elif options["imagefile"]:
image_files = options["imagefile"]
image_file_count = 1;
image_path = os.path.abspath(image_files[0])
image_split_path = os.path.split(image_path)
book_code = os.path.split(image_split_path[0])[1]
book_id = 0
if options.has_key("sql") and options["sql"]:
book_id = sql_make_book_and_return_id(book_code)
# if options.has_key("hocrout") and options["hocrout"]:
# hocr_tree = hocr_make_tree_and_return(book_code)
for image_file in image_files:
image_path = os.path.abspath(image_file)
image_split_path = os.path.split(image_path)
book_code = os.path.split(image_split_path[0])[1]#directory name
image_file_name = image_split_path[1]
imageBase, imageEx = os.path.splitext(image_file_name)
threshold_info = ""
print "Now working with image: " + image_file_name
internal_image_file_path = os.path.join(book_code, image_file_name)
if imageEx == ".jp2":
try:
jp2Image = mh.imread(image_file, as_grey=True)
jp2Image = jp2Image.astype(np.uint8)
imageIn = numpy_io.from_numpy(jp2Image)
except:
print "Unexpected error:", sys.exc_info()[0]
raise
else:
try:
imageIn = load_image(image_file)
except:
continue
imageType = imageIn.data.pixel_type
if imageType != ONEBIT:
if imageType != GREYSCALE:
imageIn = imageIn.to_greyscale()
otsu_thresh = imageIn.otsu_find_threshold()
for otsu_factor in otsu_factors:
if options.has_key("hocrout") and options["hocrout"]:
hocr_tree = hocr_make_tree_and_return(book_code)
if imageIn.data.pixel_type == ONEBIT:
threshold_info = "thresh_128"
otsu_thresh = 1.0
image = imageIn
if options["debug"]:
print "image is ONEBIT; doing no threshold optimization."
else:
current_thresh = otsu_thresh * otsu_factor
if current_thresh > 253.0:
current_thresh = 253.0
current_thresh = int(current_thresh)
threshold_info = "thresh_" + str(int(current_thresh))# + "=" + str(otsu_factor)
image = imageIn.threshold(current_thresh)
print "Otsu factor: ", otsu_factor, " threshold: ", current_thresh
if options["hocrfile"]:
hocr_to_use = string.replace(options["hocrfile"],"%s",imageBase)
g.hocr = hocr_to_use
if options["debug"]:
print "using '" + hocr_to_use + "' as hocr file"
if options.has_key("filter") and options["filter"] == True:
count = 0
ccs = image.cc_analysis()
if options.has_key("debug"):
if options["debug"] == True:
print "filter started on",len(ccs) ,"elements..."
#filter long vertical runs left over from margins
## #Agressive run filtering
## median_height = median([cc.nrows for cc in ccs])
## for cc in ccs:
## #TODO: add another condition that keeps these at edges of page
## if((cc.nrows / cc.ncols > 6) and (cc.nrows > 1.5 * median_height) ):
## cc.fill_white()
## del cc
## count = count + 1
median_black_area = median([cc.black_area()[0] for cc in ccs])
for cc in ccs:
if(cc.black_area()[0] > (median_black_area * 10)):
cc.fill_white()
del cc
count = count + 1
for cc in ccs:
if(cc.black_area()[0] < (median_black_area / 10)):
cc.fill_white()
del cc
count = count + 1
if options.has_key("debug") and options["debug"] == True:
print "filter done.",len(ccs)-count,"elements left."
if (len(ccs) < 5) or (len(ccs) > MAX_CCS):
print "Error: there are " + str(len(ccs)) + " ccs. Max is " + str( MAX_CCS) + " Omitting this image."
#raise ImageSegmentationError("Error: there are " + str(len(ccs)) + " ccs. Max is " + str( MAX_CCS) + " Omitting this image.")
else:
if options.has_key("deskew") and options["deskew"] == True:
#from gamera.toolkits.otr.otr_staff import *
if options.has_key("debug") and options["debug"] == True:
print "\ntry to skew correct..."
rotation = image.rotation_angle_projections(-10,10)[0]
img = image.rotate(rotation,0)
if options.has_key("debug") and options["debug"] == True:
print "rotated with",rotation,"angle"
if options.has_key("mode") and options["mode"] == "teubner":
(body_image, app_crit_image) = splitAppCritTeubner(image)
output = g.process_image(body_image)
if app_crit_image:
print "there is an app. crit image"
appcrit_output = g_appcrit.process_image(app_crit_image)
else:
print "there is no app. crit image"
appcrit_output = ""
output = output + appcrit_output
else:
output = g.process_image(image)
output_file_name_base = options["unicodeoutfile"] + imageBase + "_" +imageEx[1:] + "_" + threshold_info
if options.has_key("debug") and options["debug"] == True:
g.save_debug_images(output_file_name_base)
if options.has_key("mode") and options["mode"] == "teubner" and app_crit_image:
#TODO: make more general
g_appcrit.save_debug_images(output_file_name_base + "_appcrit")
if options.has_key("hocrout") and options["hocrout"]:
#if we turned this on, we would make a separate div for each page of input
#hocr_tree = hocr_make_page_and_return_div(internal_image_file_path,image_file_count,book_id,hocr_tree)
g.store_hocr(internal_image_file_path,hocr_tree)
if options.has_key("mode") and options["mode"] == "teubner" and app_crit_image:
g_appcrit.store_hocr(internal_image_file_path,hocr_tree)
if options.has_key("sql") and options["sql"]:
page_id = sql_make_page_and_return_id(internal_image_file_path,image_file_count,book_id)
g.store_sql(image_path,page_id)
if options.has_key("unicodeoutfile"):
if options.has_key("hocrout") and options["hocrout"]:
g.save_text_hocr(hocr_tree, output_file_name_base + ".html")
else:
g.save_text_unicode( output_file_name_base + ".txt")
if options.has_key("mode") and options["mode"] == "teubner":
#TODO: make the above more general
g_appcrit.save_text_unicode( output_file_name_base + "_appcrit.txt")
elif options.has_key("teubneroutfile"):
g.save_text_teubner(options["teubneroutfile"])
else:
print output
image_file_count += 1
