def __init__(self, source_image_path):
# Constants
self.LEDGER_LINE_MAX_THICKNESS = 2
self.whole_image = SheetImageSlice(Image.open(source_image_path))
#self.staffs = self.find_staffs() # list of StaffImages representing staff positions
self.notes = [] # list of Notes
self.header = None # if non-None, the main header text (usually title)
self.subheader = None # if non-None, the sub header text (usually author)
self.other_texts = [] # list of other texts found
self.ledger_dict = {} # y-values -> pitch values
self.ledger_lines = []
self.build_ledger_dict()
self.erase_ledger_lines()
self.get_notes()
def __init__(self, source_image_path):
# Constants
self.LEDGER_LINE_MAX_THICKNESS = 2
self.whole_image = SheetImageSlice(Image.open(source_image_path))
#self.staffs = self.find_staffs() # list of StaffImages representing staff positions
self.notes = [] # list of Notes
self.header = None # if non-None, the main header text (usually title)
self.subheader = None # if non-None, the sub header text (usually author)
self.other_texts = [] # list of other texts found
self.ledger_dict = {} # y-values -> pitch values
self.ledger_lines = []
self.build_ledger_dict()
self.erase_ledger_lines()
self.get_notes()
class SheetImageParser:
def __init__(self, source_image_path):
# Constants
self.LEDGER_LINE_MAX_THICKNESS = 2
self.whole_image = SheetImageSlice(Image.open(source_image_path))
#self.staffs = self.find_staffs() # list of StaffImages representing staff positions
self.notes = [] # list of Notes
self.header = None # if non-None, the main header text (usually title)
self.subheader = None # if non-None, the sub header text (usually author)
self.other_texts = [] # list of other texts found
self.ledger_dict = {} # y-values -> pitch values
self.ledger_lines = []
self.build_ledger_dict()
self.erase_ledger_lines()
self.get_notes()
def __ascii_image(self):
for y in range(self.whole_image.get_height()):
line = ""
for x in range(self.whole_image.get_width()):
line += "_" if self.whole_image.get_pixel(x, y) else "1"
def get_ledger_lines(self):
# returns a list of y-positions where there are black, horizontal lines
# we'll assume a line is anything that is 20% of the width of the image,
# is located at the center, and has a spacing of at least 3 pixels from
# the last line examined.
horizontal_lines = []
midpoint = self.whole_image.get_midpoint()
iterator_y = range(self.whole_image.get_height()).__iter__()
for y in iterator_y:
# if the center pixel is not black, it's not a line
if self.whole_image.get_pixel(midpoint, y): # pixel is white
continue
current_row = self.whole_image.get_row(y)
x_begin = int(self.whole_image.get_midpoint() - int(self.whole_image.get_width()*.1))
x_end = int(self.whole_image.get_midpoint() + int(self.whole_image.get_width()*.1))
expected_width = int(x_end-x_begin)
if current_row[x_begin:x_end].count(False) == expected_width:
horizontal_lines.append(y)
collections.deque(itertools.islice(iterator_y, self.LEDGER_LINE_MAX_THICKNESS))
self.ledger_lines = horizontal_lines
return horizontal_lines
def build_ledger_dict(self):
ledger_lines = self.get_ledger_lines()
# get distance between ledger lines
ledger_distance = ledger_lines[1] - ledger_lines[0]
half_distance = int(ledger_distance/2)
for i, line in enumerate(ledger_lines):
if (i+1) % 5 == 1:
self.ledger_dict[line - half_distance] = 'G3'
self.ledger_dict[line] = 'F3'
self.ledger_dict[line + half_distance] = 'E3'
elif (i+1) % 5 == 2:
self.ledger_dict[line] = 'D3'
self.ledger_dict[line + half_distance] = 'C3'
elif (i+1) % 5 == 3:
self.ledger_dict[line] = 'B3'
self.ledger_dict[line + half_distance] = 'A4'
elif (i+1) % 5 == 4:
self.ledger_dict[line] = 'G4'
self.ledger_dict[line + half_distance] = 'F4'
elif (i+1) % 5 == 0:
self.ledger_dict[line] = 'E4'
self.ledger_dict[line + half_distance] = 'D4'
self.ledger_dict[line + half_distance*2] = 'C4'
def erase_ledger_lines(self):
# how many are there
counts = []
for x in range(self.whole_image.get_width()):
count = 0
for y in range(self.whole_image.get_height()):
if not self.whole_image.get_pixel(x, y):
count += 1
counts.append(count)
total = collections.Counter(counts).most_common(1)[0][0]
# where are they
locations = []
for x in range(self.whole_image.get_width()):
count = 0
blacks_found = []
for y in range(self.whole_image.get_height()):
if not self.whole_image.get_pixel(x, y):
count += 1
blacks_found.append(y)
if count == total:
locations = blacks_found
break
# erase
for y in locations:
for x in range(self.whole_image.get_width()):
if self.whole_image.get_pixel(x, y-2) or self.whole_image.get_pixel(x, y+2):
self.whole_image.erase(x, y)
# testing
self.whole_image.image.save("test.gif")
def get_notes(self):
# returns list of note images
checked = [] # list of rectangle tuples of areas we've already examined
notes = [] # list of NoteImage notes
ledgers = sorted(self.ledger_lines)
top_y = ledgers[0]
bottom_y = ledgers[-1]
left_x = 3
right_x = self.whole_image.get_width()-3
head_size = ledgers[1] - ledgers[0]
imaginary_borders = []
begin = None
end = None
for i, y in enumerate(ledgers):
if (i+1) % 5 == 1:
begin = y
if (i+1) % 5 == 0:
end = y
distance = end - begin
buffer = distance/2
imaginary_borders.append((begin-buffer, end+buffer))
for borders in imaginary_borders:
for x in range(left_x, right_x):
for y in range(borders[0], borders[1]):
# make sure we didn't check this pixel already
already_checked = False
for rect in checked:
if x >= rect[0] and x <= rect[0]+rect[2]:
if y >= rect[1] and y <= rect[1] + rect[3]:
already_checked = True
break
if already_checked:
continue
if not self.whole_image.get_pixel(x, y):
# begin boxing out note
top_left = {"x": x, "y": y}
bottom_right = {"x": x+1, "y": y+1}
search_over = False
while not search_over:
search_over = True
# look up
for xi in range(top_left["x"]-1, bottom_right["x"]+1):
if not self.whole_image.get_pixel(xi, top_left["y"]-1):
search_over = False
top_left["y"] -= 1
break
# look down
for xi in range(top_left["x"]-1, bottom_right["x"]+1):
if not self.whole_image.get_pixel(xi, bottom_right["y"]+1):
search_over = False
bottom_right["y"] += 1
break
# look left
for yi in range(top_left["y"]-1, bottom_right["y"]+1):
if not self.whole_image.get_pixel(top_left["x"]-1, yi):
search_over = False
top_left["x"] -= 1
break
# look right
for yi in range(top_left["y"]-1, bottom_right["y"]+1):
if not self.whole_image.get_pixel(bottom_right["x"]+1, yi):
search_over = False
bottom_right["x"] += 1
break
# search is over
note_x_pos = top_left["x"]
note_y_pos = top_left["y"]
note_width = abs(bottom_right["x"] - top_left["x"])
note_height = abs(bottom_right["y"] - top_left["y"])
checked.append([note_x_pos, note_y_pos, note_width, note_height])
if note_width > 3 and note_height > 3:
note = NoteImage(self.whole_image.slice(note_x_pos, note_y_pos, note_width+1, note_height+1),
head_size, note_x_pos + note_width, note_y_pos + note_height - head_size/2)
notes.append(note)
self.notes = notes
def get_pitch(self, y):
# get closest line
lines = self.ledger_dict.keys()
smallest_difference = None
note = None
for line in lines:
diff = abs(line - y)
if smallest_difference == None or diff < smallest_difference:
smallest_difference = diff
note = self.ledger_dict[line]
return note
class SheetImageParser:
def __init__(self, source_image_path):
# Constants
self.LEDGER_LINE_MAX_THICKNESS = 2
self.whole_image = SheetImageSlice(Image.open(source_image_path))
#self.staffs = self.find_staffs() # list of StaffImages representing staff positions
self.notes = [] # list of Notes
self.header = None # if non-None, the main header text (usually title)
self.subheader = None # if non-None, the sub header text (usually author)
self.other_texts = [] # list of other texts found
self.ledger_dict = {} # y-values -> pitch values
self.ledger_lines = []
self.build_ledger_dict()
self.erase_ledger_lines()
self.get_notes()
def __ascii_image(self):
for y in range(self.whole_image.get_height()):
line = ""
for x in range(self.whole_image.get_width()):
line += "_" if self.whole_image.get_pixel(x, y) else "1"
def get_ledger_lines(self):
# returns a list of y-positions where there are black, horizontal lines
# we'll assume a line is anything that is 20% of the width of the image,
# is located at the center, and has a spacing of at least 3 pixels from
# the last line examined.
horizontal_lines = []
midpoint = self.whole_image.get_midpoint()
iterator_y = range(self.whole_image.get_height()).__iter__()
for y in iterator_y:
# if the center pixel is not black, it's not a line
if self.whole_image.get_pixel(midpoint, y): # pixel is white
continue
current_row = self.whole_image.get_row(y)
x_begin = int(self.whole_image.get_midpoint() -
int(self.whole_image.get_width() * .1))
x_end = int(self.whole_image.get_midpoint() +
int(self.whole_image.get_width() * .1))
expected_width = int(x_end - x_begin)
if current_row[x_begin:x_end].count(False) == expected_width:
horizontal_lines.append(y)
collections.deque(
itertools.islice(iterator_y,
self.LEDGER_LINE_MAX_THICKNESS))
self.ledger_lines = horizontal_lines
return horizontal_lines
def build_ledger_dict(self):
ledger_lines = self.get_ledger_lines()
# get distance between ledger lines
ledger_distance = ledger_lines[1] - ledger_lines[0]
half_distance = int(ledger_distance / 2)
for i, line in enumerate(ledger_lines):
if (i + 1) % 5 == 1:
self.ledger_dict[line - half_distance] = 'G3'
self.ledger_dict[line] = 'F3'
self.ledger_dict[line + half_distance] = 'E3'
elif (i + 1) % 5 == 2:
self.ledger_dict[line] = 'D3'
self.ledger_dict[line + half_distance] = 'C3'
elif (i + 1) % 5 == 3:
self.ledger_dict[line] = 'B3'
self.ledger_dict[line + half_distance] = 'A4'
elif (i + 1) % 5 == 4:
self.ledger_dict[line] = 'G4'
self.ledger_dict[line + half_distance] = 'F4'
elif (i + 1) % 5 == 0:
self.ledger_dict[line] = 'E4'
self.ledger_dict[line + half_distance] = 'D4'
self.ledger_dict[line + half_distance * 2] = 'C4'
def erase_ledger_lines(self):
# how many are there
counts = []
for x in range(self.whole_image.get_width()):
count = 0
for y in range(self.whole_image.get_height()):
if not self.whole_image.get_pixel(x, y):
count += 1
counts.append(count)
total = collections.Counter(counts).most_common(1)[0][0]
# where are they
locations = []
for x in range(self.whole_image.get_width()):
count = 0
blacks_found = []
for y in range(self.whole_image.get_height()):
if not self.whole_image.get_pixel(x, y):
count += 1
blacks_found.append(y)
if count == total:
locations = blacks_found
break
# erase
for y in locations:
for x in range(self.whole_image.get_width()):
if self.whole_image.get_pixel(
x, y - 2) or self.whole_image.get_pixel(x, y + 2):
self.whole_image.erase(x, y)
# testing
self.whole_image.image.save("test.gif")
def get_notes(self):
# returns list of note images
checked = [
] # list of rectangle tuples of areas we've already examined
notes = [] # list of NoteImage notes
ledgers = sorted(self.ledger_lines)
top_y = ledgers[0]
bottom_y = ledgers[-1]
left_x = 3
right_x = self.whole_image.get_width() - 3
head_size = ledgers[1] - ledgers[0]
imaginary_borders = []
begin = None
end = None
for i, y in enumerate(ledgers):
if (i + 1) % 5 == 1:
begin = y
if (i + 1) % 5 == 0:
end = y
distance = end - begin
buffer = distance / 2
imaginary_borders.append((begin - buffer, end + buffer))
for borders in imaginary_borders:
for x in range(left_x, right_x):
for y in range(borders[0], borders[1]):
# make sure we didn't check this pixel already
already_checked = False
for rect in checked:
if x >= rect[0] and x <= rect[0] + rect[2]:
if y >= rect[1] and y <= rect[1] + rect[3]:
already_checked = True
break
if already_checked:
continue
if not self.whole_image.get_pixel(x, y):
# begin boxing out note
top_left = {"x": x, "y": y}
bottom_right = {"x": x + 1, "y": y + 1}
search_over = False
while not search_over:
search_over = True
# look up
for xi in range(top_left["x"] - 1,
bottom_right["x"] + 1):
if not self.whole_image.get_pixel(
xi, top_left["y"] - 1):
search_over = False
top_left["y"] -= 1
break
# look down
for xi in range(top_left["x"] - 1,
bottom_right["x"] + 1):
if not self.whole_image.get_pixel(
xi, bottom_right["y"] + 1):
search_over = False
bottom_right["y"] += 1
break
# look left
for yi in range(top_left["y"] - 1,
bottom_right["y"] + 1):
if not self.whole_image.get_pixel(
top_left["x"] - 1, yi):
search_over = False
top_left["x"] -= 1
break
# look right
for yi in range(top_left["y"] - 1,
bottom_right["y"] + 1):
if not self.whole_image.get_pixel(
bottom_right["x"] + 1, yi):
search_over = False
bottom_right["x"] += 1
break
# search is over
note_x_pos = top_left["x"]
note_y_pos = top_left["y"]
note_width = abs(bottom_right["x"] - top_left["x"])
note_height = abs(bottom_right["y"] - top_left["y"])
checked.append(
[note_x_pos, note_y_pos, note_width, note_height])
if note_width > 3 and note_height > 3:
note = NoteImage(
self.whole_image.slice(note_x_pos, note_y_pos,
note_width + 1,
note_height + 1),
head_size, note_x_pos + note_width,
note_y_pos + note_height - head_size / 2)
notes.append(note)
self.notes = notes
def get_pitch(self, y):
# get closest line
lines = self.ledger_dict.keys()
smallest_difference = None
note = None
for line in lines:
diff = abs(line - y)
if smallest_difference == None or diff < smallest_difference:
smallest_difference = diff
note = self.ledger_dict[line]
return note
