mirror of
https://github.com/LmeSzinc/StarRailCopilot.git
synced 2024-11-23 00:52:22 +00:00
934 lines
24 KiB
Python
934 lines
24 KiB
Python
import re
|
|
|
|
import cv2
|
|
import numpy as np
|
|
from PIL import Image
|
|
|
|
REGEX_NODE = re.compile(r'(-?[A-Za-z]+)(-?\d+)')
|
|
|
|
|
|
def random_normal_distribution_int(a, b, n=3):
|
|
"""Generate a normal distribution int within the interval. Use the average value of several random numbers to
|
|
simulate normal distribution.
|
|
|
|
Args:
|
|
a (int): The minimum of the interval.
|
|
b (int): The maximum of the interval.
|
|
n (int): The amount of numbers in simulation. Default to 3.
|
|
|
|
Returns:
|
|
int
|
|
"""
|
|
if a < b:
|
|
output = np.mean(np.random.randint(a, b, size=n))
|
|
return int(output.round())
|
|
else:
|
|
return b
|
|
|
|
|
|
def random_rectangle_point(area, n=3):
|
|
"""Choose a random point in an area.
|
|
|
|
Args:
|
|
area: (upper_left_x, upper_left_y, bottom_right_x, bottom_right_y).
|
|
n (int): The amount of numbers in simulation. Default to 3.
|
|
|
|
Returns:
|
|
tuple(int): (x, y)
|
|
"""
|
|
x = random_normal_distribution_int(area[0], area[2], n=n)
|
|
y = random_normal_distribution_int(area[1], area[3], n=n)
|
|
return x, y
|
|
|
|
|
|
def random_rectangle_vector(vector, box, random_range=(0, 0, 0, 0), padding=15):
|
|
"""Place a vector in a box randomly.
|
|
|
|
Args:
|
|
vector: (x, y)
|
|
box: (upper_left_x, upper_left_y, bottom_right_x, bottom_right_y).
|
|
random_range (tuple): Add a random_range to vector. (x_min, y_min, x_max, y_max).
|
|
padding (int):
|
|
|
|
Returns:
|
|
tuple(int), tuple(int): start_point, end_point.
|
|
"""
|
|
vector = np.array(vector) + random_rectangle_point(random_range)
|
|
vector = np.round(vector).astype(int)
|
|
half_vector = np.round(vector / 2).astype(int)
|
|
box = np.array(box) + np.append(np.abs(half_vector) + padding, -np.abs(half_vector) - padding)
|
|
center = random_rectangle_point(box)
|
|
start_point = center - half_vector
|
|
end_point = start_point + vector
|
|
return tuple(start_point), tuple(end_point)
|
|
|
|
|
|
def random_rectangle_vector_opted(
|
|
vector, box, random_range=(0, 0, 0, 0), padding=15, whitelist_area=None, blacklist_area=None):
|
|
"""
|
|
Place a vector in a box randomly.
|
|
|
|
When emulator/game stuck, it treats a swipe as a click, clicking at the end of swipe path.
|
|
To prevent this, random results need to be filtered.
|
|
|
|
Args:
|
|
vector: (x, y)
|
|
box: (upper_left_x, upper_left_y, bottom_right_x, bottom_right_y).
|
|
random_range (tuple): Add a random_range to vector. (x_min, y_min, x_max, y_max).
|
|
padding (int):
|
|
whitelist_area: (list[tuple[int]]):
|
|
A list of area that safe to click. Swipe path will end there.
|
|
blacklist_area: (list[tuple[int]]):
|
|
If none of the whitelist_area satisfies current vector, blacklist_area will be used.
|
|
Delete random path that ends in any blacklist_area.
|
|
|
|
Returns:
|
|
tuple(int), tuple(int): start_point, end_point.
|
|
"""
|
|
vector = np.array(vector) + random_rectangle_point(random_range)
|
|
vector = np.round(vector).astype(int)
|
|
half_vector = np.round(vector / 2).astype(int)
|
|
box_pad = np.array(box) + np.append(np.abs(half_vector) + padding, -np.abs(half_vector) - padding)
|
|
box_pad = area_offset(box_pad, half_vector)
|
|
segment = int(np.linalg.norm(vector) // 70) + 1
|
|
|
|
def in_blacklist(end):
|
|
if not blacklist_area:
|
|
return False
|
|
for x in range(segment + 1):
|
|
point = - vector * x / segment + end
|
|
for area in blacklist_area:
|
|
if point_in_area(point, area, threshold=0):
|
|
return True
|
|
return False
|
|
|
|
if whitelist_area:
|
|
for area in whitelist_area:
|
|
area = area_limit(area, box_pad)
|
|
if all([x > 0 for x in area_size(area)]):
|
|
end_point = random_rectangle_point(area)
|
|
for _ in range(10):
|
|
if in_blacklist(end_point):
|
|
continue
|
|
return point_limit(end_point - vector, box), point_limit(end_point, box)
|
|
|
|
for _ in range(100):
|
|
end_point = random_rectangle_point(box_pad)
|
|
if in_blacklist(end_point):
|
|
continue
|
|
return point_limit(end_point - vector, box), point_limit(end_point, box)
|
|
|
|
end_point = random_rectangle_point(box_pad)
|
|
return point_limit(end_point - vector, box), point_limit(end_point, box)
|
|
|
|
|
|
def random_line_segments(p1, p2, n, random_range=(0, 0, 0, 0)):
|
|
"""Cut a line into multiple segments.
|
|
|
|
Args:
|
|
p1: (x, y).
|
|
p2: (x, y).
|
|
n: Number of slice.
|
|
random_range: Add a random_range to points.
|
|
|
|
Returns:
|
|
list[tuple]: [(x0, y0), (x1, y1), (x2, y2)]
|
|
"""
|
|
return [tuple((((n - index) * p1 + index * p2) / n).astype(int) + random_rectangle_point(random_range))
|
|
for index in range(0, n + 1)]
|
|
|
|
|
|
def ensure_time(second, n=3, precision=3):
|
|
"""Ensure to be time.
|
|
|
|
Args:
|
|
second (int, float, tuple): time, such as 10, (10, 30), '10, 30'
|
|
n (int): The amount of numbers in simulation. Default to 5.
|
|
precision (int): Decimals.
|
|
|
|
Returns:
|
|
float:
|
|
"""
|
|
if isinstance(second, tuple):
|
|
multiply = 10 ** precision
|
|
result = random_normal_distribution_int(second[0] * multiply, second[1] * multiply, n) / multiply
|
|
return round(result, precision)
|
|
elif isinstance(second, str):
|
|
if ',' in second:
|
|
lower, upper = second.replace(' ', '').split(',')
|
|
lower, upper = int(lower), int(upper)
|
|
return ensure_time((lower, upper), n=n, precision=precision)
|
|
if '-' in second:
|
|
lower, upper = second.replace(' ', '').split('-')
|
|
lower, upper = int(lower), int(upper)
|
|
return ensure_time((lower, upper), n=n, precision=precision)
|
|
else:
|
|
return int(second)
|
|
else:
|
|
return second
|
|
|
|
|
|
def ensure_int(*args):
|
|
"""
|
|
Convert all elements to int.
|
|
Return the same structure as nested objects.
|
|
|
|
Args:
|
|
*args:
|
|
|
|
Returns:
|
|
list:
|
|
"""
|
|
|
|
def to_int(item):
|
|
try:
|
|
return int(item)
|
|
except TypeError:
|
|
result = [to_int(i) for i in item]
|
|
if len(result) == 1:
|
|
result = result[0]
|
|
return result
|
|
|
|
return to_int(args)
|
|
|
|
|
|
def area_offset(area, offset):
|
|
"""
|
|
Move an area.
|
|
|
|
Args:
|
|
area: (upper_left_x, upper_left_y, bottom_right_x, bottom_right_y).
|
|
offset: (x, y).
|
|
|
|
Returns:
|
|
tuple: (upper_left_x, upper_left_y, bottom_right_x, bottom_right_y).
|
|
"""
|
|
upper_left_x, upper_left_y, bottom_right_x, bottom_right_y = area
|
|
x, y = offset
|
|
return upper_left_x + x, upper_left_y + y, bottom_right_x + x, bottom_right_y + y
|
|
|
|
|
|
def area_pad(area, pad=10):
|
|
"""
|
|
Inner offset an area.
|
|
|
|
Args:
|
|
area: (upper_left_x, upper_left_y, bottom_right_x, bottom_right_y).
|
|
pad (int):
|
|
|
|
Returns:
|
|
tuple: (upper_left_x, upper_left_y, bottom_right_x, bottom_right_y).
|
|
"""
|
|
upper_left_x, upper_left_y, bottom_right_x, bottom_right_y = area
|
|
return upper_left_x + pad, upper_left_y + pad, bottom_right_x - pad, bottom_right_y - pad
|
|
|
|
|
|
def limit_in(x, lower, upper):
|
|
"""
|
|
Limit x within range (lower, upper)
|
|
|
|
Args:
|
|
x:
|
|
lower:
|
|
upper:
|
|
|
|
Returns:
|
|
int, float:
|
|
"""
|
|
return max(min(x, upper), lower)
|
|
|
|
|
|
def area_limit(area1, area2):
|
|
"""
|
|
Limit an area in another area.
|
|
|
|
Args:
|
|
area1: (upper_left_x, upper_left_y, bottom_right_x, bottom_right_y).
|
|
area2: (upper_left_x, upper_left_y, bottom_right_x, bottom_right_y).
|
|
|
|
Returns:
|
|
tuple: (upper_left_x, upper_left_y, bottom_right_x, bottom_right_y).
|
|
"""
|
|
x_lower, y_lower, x_upper, y_upper = area2
|
|
return (
|
|
limit_in(area1[0], x_lower, x_upper),
|
|
limit_in(area1[1], y_lower, y_upper),
|
|
limit_in(area1[2], x_lower, x_upper),
|
|
limit_in(area1[3], y_lower, y_upper),
|
|
)
|
|
|
|
|
|
def area_size(area):
|
|
"""
|
|
Area size or shape.
|
|
|
|
Args:
|
|
area: (upper_left_x, upper_left_y, bottom_right_x, bottom_right_y).
|
|
|
|
Returns:
|
|
tuple: (x, y).
|
|
"""
|
|
return (
|
|
max(area[2] - area[0], 0),
|
|
max(area[3] - area[1], 0)
|
|
)
|
|
|
|
|
|
def area_center(area):
|
|
"""
|
|
Get the center of an area
|
|
|
|
Args:
|
|
area: (upper_left_x, upper_left_y, bottom_right_x, bottom_right_y)
|
|
|
|
Returns:
|
|
tuple: (x, y)
|
|
"""
|
|
x1, y1, x2, y2 = area
|
|
return (x1 + x2) / 2, (y1 + y2) / 2
|
|
|
|
|
|
def point_limit(point, area):
|
|
"""
|
|
Limit point in an area.
|
|
|
|
Args:
|
|
point: (x, y).
|
|
area: (upper_left_x, upper_left_y, bottom_right_x, bottom_right_y).
|
|
|
|
Returns:
|
|
tuple: (x, y).
|
|
"""
|
|
return (
|
|
limit_in(point[0], area[0], area[2]),
|
|
limit_in(point[1], area[1], area[3])
|
|
)
|
|
|
|
|
|
def point_in_area(point, area, threshold=5):
|
|
"""
|
|
|
|
Args:
|
|
point: (x, y).
|
|
area: (upper_left_x, upper_left_y, bottom_right_x, bottom_right_y).
|
|
threshold: int
|
|
|
|
Returns:
|
|
bool:
|
|
"""
|
|
return area[0] - threshold < point[0] < area[2] + threshold and area[1] - threshold < point[1] < area[3] + threshold
|
|
|
|
|
|
def area_in_area(area1, area2, threshold=5):
|
|
"""
|
|
|
|
Args:
|
|
area1: (upper_left_x, upper_left_y, bottom_right_x, bottom_right_y).
|
|
area2: (upper_left_x, upper_left_y, bottom_right_x, bottom_right_y).
|
|
threshold: int
|
|
|
|
Returns:
|
|
bool:
|
|
"""
|
|
return area2[0] - threshold <= area1[0] \
|
|
and area2[1] - threshold <= area1[1] \
|
|
and area1[2] <= area2[2] + threshold \
|
|
and area1[3] <= area2[3] + threshold
|
|
|
|
|
|
def area_cross_area(area1, area2, threshold=5):
|
|
"""
|
|
|
|
Args:
|
|
area1: (upper_left_x, upper_left_y, bottom_right_x, bottom_right_y).
|
|
area2: (upper_left_x, upper_left_y, bottom_right_x, bottom_right_y).
|
|
threshold: int
|
|
|
|
Returns:
|
|
bool:
|
|
"""
|
|
# https://www.yiiven.cn/rect-is-intersection.html
|
|
xa1, ya1, xa2, ya2 = area1
|
|
xb1, yb1, xb2, yb2 = area2
|
|
return abs(xb2 + xb1 - xa2 - xa1) <= xa2 - xa1 + xb2 - xb1 + threshold * 2 \
|
|
and abs(yb2 + yb1 - ya2 - ya1) <= ya2 - ya1 + yb2 - yb1 + threshold * 2
|
|
|
|
|
|
def float2str(n, decimal=3):
|
|
"""
|
|
Args:
|
|
n (float):
|
|
decimal (int):
|
|
|
|
Returns:
|
|
str:
|
|
"""
|
|
return str(round(n, decimal)).ljust(decimal + 2, "0")
|
|
|
|
|
|
def point2str(x, y, length=4):
|
|
"""
|
|
Args:
|
|
x (int, float):
|
|
y (int, float):
|
|
length (int): Align length.
|
|
|
|
Returns:
|
|
str: String with numbers right aligned, such as '( 100, 80)'.
|
|
"""
|
|
return '(%s, %s)' % (str(int(x)).rjust(length), str(int(y)).rjust(length))
|
|
|
|
|
|
def col2name(col):
|
|
"""
|
|
Convert a zero indexed column cell reference to a string.
|
|
|
|
Args:
|
|
col: The cell column. Int.
|
|
|
|
Returns:
|
|
Column style string.
|
|
|
|
Examples:
|
|
0 -> A, 3 -> D, 35 -> AJ, -1 -> -A
|
|
"""
|
|
|
|
col_neg = col < 0
|
|
if col_neg:
|
|
col_num = -col
|
|
else:
|
|
col_num = col + 1 # Change to 1-index.
|
|
col_str = ''
|
|
|
|
while col_num:
|
|
# Set remainder from 1 .. 26
|
|
remainder = col_num % 26
|
|
|
|
if remainder == 0:
|
|
remainder = 26
|
|
|
|
# Convert the remainder to a character.
|
|
col_letter = chr(remainder + 64)
|
|
|
|
# Accumulate the column letters, right to left.
|
|
col_str = col_letter + col_str
|
|
|
|
# Get the next order of magnitude.
|
|
col_num = int((col_num - 1) / 26)
|
|
|
|
if col_neg:
|
|
return '-' + col_str
|
|
else:
|
|
return col_str
|
|
|
|
|
|
def name2col(col_str):
|
|
"""
|
|
Convert a cell reference in A1 notation to a zero indexed row and column.
|
|
|
|
Args:
|
|
col_str: A1 style string.
|
|
|
|
Returns:
|
|
row, col: Zero indexed cell row and column indices.
|
|
"""
|
|
# Convert base26 column string to number.
|
|
expn = 0
|
|
col = 0
|
|
col_neg = col_str.startswith('-')
|
|
col_str = col_str.strip('-').upper()
|
|
|
|
for char in reversed(col_str):
|
|
col += (ord(char) - 64) * (26 ** expn)
|
|
expn += 1
|
|
|
|
if col_neg:
|
|
return -col
|
|
else:
|
|
return col - 1 # Convert 1-index to zero-index
|
|
|
|
|
|
def node2location(node):
|
|
"""
|
|
See location2node()
|
|
|
|
Args:
|
|
node (str): Example: 'E3'
|
|
|
|
Returns:
|
|
tuple[int]: Example: (4, 2)
|
|
"""
|
|
res = REGEX_NODE.search(node)
|
|
if res:
|
|
x, y = res.group(1), res.group(2)
|
|
y = int(y)
|
|
if y > 0:
|
|
y -= 1
|
|
return name2col(x), y
|
|
else:
|
|
# Whatever
|
|
return ord(node[0]) % 32 - 1, int(node[1:]) - 1
|
|
|
|
|
|
def location2node(location):
|
|
"""
|
|
Convert location tuple to an Excel-like cell.
|
|
Accept negative values also.
|
|
|
|
-2 -1 0 1 2 3
|
|
-2 -B-2 -A-2 A-2 B-2 C-2 D-2
|
|
-1 -B-1 -A-1 A-1 B-1 C-1 D-1
|
|
0 -B1 -A1 A1 B1 C1 D1
|
|
1 -B2 -A2 A2 B2 C2 D2
|
|
2 -B3 -A3 A3 B3 C3 D3
|
|
3 -B4 -A4 A4 B4 C4 D4
|
|
|
|
# To generate the table above
|
|
index = range(-2, 4)
|
|
row = ' ' + ' '.join([str(i).rjust(4) for i in index])
|
|
print(row)
|
|
for y in index:
|
|
row = str(y).rjust(2) + ' ' + ' '.join([location2node((x, y)).rjust(4) for x in index])
|
|
print(row)
|
|
|
|
def check(node):
|
|
return point2str(*node2location(location2node(node)), length=2)
|
|
row = ' ' + ' '.join([str(i).rjust(8) for i in index])
|
|
print(row)
|
|
for y in index:
|
|
row = str(y).rjust(2) + ' ' + ' '.join([check((x, y)).rjust(4) for x in index])
|
|
print(row)
|
|
|
|
Args:
|
|
location (tuple[int]):
|
|
|
|
Returns:
|
|
str:
|
|
"""
|
|
x, y = location
|
|
if y >= 0:
|
|
y += 1
|
|
return col2name(x) + str(y)
|
|
|
|
|
|
def load_image(file, area=None):
|
|
"""
|
|
Load an image like pillow and drop alpha channel.
|
|
|
|
Args:
|
|
file (str):
|
|
area (tuple):
|
|
|
|
Returns:
|
|
np.ndarray:
|
|
"""
|
|
image = Image.open(file)
|
|
if area is not None:
|
|
image = image.crop(area)
|
|
image = np.array(image)
|
|
channel = image.shape[2] if len(image.shape) > 2 else 1
|
|
if channel > 3:
|
|
image = image[:, :, :3].copy()
|
|
return image
|
|
|
|
|
|
def save_image(image, file):
|
|
"""
|
|
Save an image like pillow.
|
|
|
|
Args:
|
|
image (np.ndarray):
|
|
file (str):
|
|
"""
|
|
# image = cv2.cvtColor(image, cv2.COLOR_RGB2BGR)
|
|
# cv2.imwrite(file, image)
|
|
Image.fromarray(image).save(file)
|
|
|
|
|
|
def crop(image, area, copy=True):
|
|
"""
|
|
Crop image like pillow, when using opencv / numpy.
|
|
Provides a black background if cropping outside of image.
|
|
|
|
Args:
|
|
image (np.ndarray):
|
|
area:
|
|
copy (bool):
|
|
|
|
Returns:
|
|
np.ndarray:
|
|
"""
|
|
x1, y1, x2, y2 = map(int, map(round, area))
|
|
h, w = image.shape[:2]
|
|
border = np.maximum((0 - y1, y2 - h, 0 - x1, x2 - w), 0)
|
|
x1, y1, x2, y2 = np.maximum((x1, y1, x2, y2), 0)
|
|
image = image[y1:y2, x1:x2]
|
|
if sum(border) > 0:
|
|
image = cv2.copyMakeBorder(image, *border, borderType=cv2.BORDER_CONSTANT, value=(0, 0, 0))
|
|
if copy:
|
|
image = image.copy()
|
|
return image
|
|
|
|
|
|
def resize(image, size):
|
|
"""
|
|
Resize image like pillow image.resize(), but implement in opencv.
|
|
Pillow uses PIL.Image.NEAREST by default.
|
|
|
|
Args:
|
|
image (np.ndarray):
|
|
size: (x, y)
|
|
|
|
Returns:
|
|
np.ndarray:
|
|
"""
|
|
return cv2.resize(image, size, interpolation=cv2.INTER_NEAREST)
|
|
|
|
|
|
def image_channel(image):
|
|
"""
|
|
Args:
|
|
image (np.ndarray):
|
|
|
|
Returns:
|
|
int: 0 for grayscale, 3 for RGB.
|
|
"""
|
|
return image.shape[2] if len(image.shape) == 3 else 0
|
|
|
|
|
|
def image_size(image):
|
|
"""
|
|
Args:
|
|
image (np.ndarray):
|
|
|
|
Returns:
|
|
int, int: width, height
|
|
"""
|
|
shape = image.shape
|
|
return shape[1], shape[0]
|
|
|
|
|
|
def image_paste(image, background, origin):
|
|
"""
|
|
Paste an image on background.
|
|
This method does not return a value, but instead updates the array "background".
|
|
|
|
Args:
|
|
image:
|
|
background:
|
|
origin: Upper-left corner, (x, y)
|
|
"""
|
|
x, y = origin
|
|
w, h = image_size(image)
|
|
background[y:y + h, x:x + w] = image
|
|
|
|
|
|
def rgb2gray(image):
|
|
"""
|
|
Args:
|
|
image (np.ndarray): Shape (height, width, channel)
|
|
|
|
Returns:
|
|
np.ndarray: Shape (height, width)
|
|
"""
|
|
r, g, b = cv2.split(image)
|
|
return cv2.add(
|
|
cv2.multiply(cv2.max(cv2.max(r, g), b), 0.5),
|
|
cv2.multiply(cv2.min(cv2.min(r, g), b), 0.5)
|
|
)
|
|
|
|
|
|
def rgb2hsv(image):
|
|
"""
|
|
Convert RGB color space to HSV color space.
|
|
HSV is Hue Saturation Value.
|
|
|
|
Args:
|
|
image (np.ndarray): Shape (height, width, channel)
|
|
|
|
Returns:
|
|
np.ndarray: Hue (0~360), Saturation (0~100), Value (0~100).
|
|
"""
|
|
image = cv2.cvtColor(image, cv2.COLOR_RGB2HSV).astype(float)
|
|
image *= (360 / 180, 100 / 255, 100 / 255)
|
|
return image
|
|
|
|
|
|
def rgb2yuv(image):
|
|
"""
|
|
Convert RGB to YUV color space.
|
|
|
|
Args:
|
|
image (np.ndarray): Shape (height, width, channel)
|
|
|
|
Returns:
|
|
np.ndarray: Shape (height, width)
|
|
"""
|
|
image = cv2.cvtColor(image, cv2.COLOR_RGB2YUV)
|
|
return image
|
|
|
|
|
|
def rgb2luma(image):
|
|
"""
|
|
Convert RGB to the Y channel (Luminance) in YUV color space.
|
|
|
|
Args:
|
|
image (np.ndarray): Shape (height, width, channel)
|
|
|
|
Returns:
|
|
np.ndarray: Shape (height, width)
|
|
"""
|
|
image = cv2.cvtColor(image, cv2.COLOR_RGB2YUV)
|
|
luma, _, _ = cv2.split(image)
|
|
return luma
|
|
|
|
|
|
def get_color(image, area):
|
|
"""Calculate the average color of a particular area of the image.
|
|
|
|
Args:
|
|
image (np.ndarray): Screenshot.
|
|
area (tuple): (upper_left_x, upper_left_y, bottom_right_x, bottom_right_y)
|
|
|
|
Returns:
|
|
tuple: (r, g, b)
|
|
"""
|
|
temp = crop(image, area, copy=False)
|
|
color = cv2.mean(temp)
|
|
return color[:3]
|
|
|
|
|
|
def get_bbox(image, threshold=0):
|
|
"""
|
|
A numpy implementation of the getbbox() in pillow.
|
|
|
|
Args:
|
|
image (np.ndarray): Screenshot.
|
|
threshold (int): Color <= threshold will be considered black
|
|
|
|
Returns:
|
|
tuple: (upper_left_x, upper_left_y, bottom_right_x, bottom_right_y)
|
|
"""
|
|
if image_channel(image) == 3:
|
|
image = np.max(image, axis=2)
|
|
x = np.where(np.max(image, axis=0) > threshold)[0]
|
|
y = np.where(np.max(image, axis=1) > threshold)[0]
|
|
return x[0], y[0], x[-1] + 1, y[-1] + 1
|
|
|
|
|
|
def get_bbox_reversed(image, threshold=0):
|
|
"""
|
|
Similar to `get_bbox` but for black contents on white background.
|
|
|
|
Args:
|
|
image (np.ndarray): Screenshot.
|
|
threshold (int): Color >= threshold will be considered white
|
|
|
|
Returns:
|
|
tuple: (upper_left_x, upper_left_y, bottom_right_x, bottom_right_y)
|
|
"""
|
|
if image_channel(image) == 3:
|
|
image = np.min(image, axis=2)
|
|
x = np.where(np.min(image, axis=0) < threshold)[0]
|
|
y = np.where(np.min(image, axis=1) < threshold)[0]
|
|
return x[0], y[0], x[-1] + 1, y[-1] + 1
|
|
|
|
|
|
def color_similarity(color1, color2):
|
|
"""
|
|
Args:
|
|
color1 (tuple): (r, g, b)
|
|
color2 (tuple): (r, g, b)
|
|
|
|
Returns:
|
|
int:
|
|
"""
|
|
diff = np.array(color1).astype(int) - np.array(color2).astype(int)
|
|
diff = np.max(np.maximum(diff, 0)) - np.min(np.minimum(diff, 0))
|
|
return diff
|
|
|
|
|
|
def color_similar(color1, color2, threshold=10):
|
|
"""Consider two colors are similar, if tolerance lesser or equal threshold.
|
|
Tolerance = Max(Positive(difference_rgb)) + Max(- Negative(difference_rgb))
|
|
The same as the tolerance in Photoshop.
|
|
|
|
Args:
|
|
color1 (tuple): (r, g, b)
|
|
color2 (tuple): (r, g, b)
|
|
threshold (int): Default to 10.
|
|
|
|
Returns:
|
|
bool: True if two colors are similar.
|
|
"""
|
|
# print(color1, color2)
|
|
diff = np.array(color1).astype(int) - np.array(color2).astype(int)
|
|
diff = np.max(np.maximum(diff, 0)) - np.min(np.minimum(diff, 0))
|
|
return diff <= threshold
|
|
|
|
|
|
def color_similar_1d(image, color, threshold=10):
|
|
"""
|
|
Args:
|
|
image (np.ndarray): 1D array.
|
|
color: (r, g, b)
|
|
threshold(int): Default to 10.
|
|
|
|
Returns:
|
|
np.ndarray: bool
|
|
"""
|
|
diff = image.astype(int) - color
|
|
diff = np.max(np.maximum(diff, 0), axis=1) - np.min(np.minimum(diff, 0), axis=1)
|
|
return diff <= threshold
|
|
|
|
|
|
def color_similarity_2d(image, color):
|
|
"""
|
|
Args:
|
|
image: 2D array.
|
|
color: (r, g, b)
|
|
|
|
Returns:
|
|
np.ndarray: uint8
|
|
"""
|
|
r, g, b = cv2.split(cv2.subtract(image, (*color, 0)))
|
|
positive = cv2.max(cv2.max(r, g), b)
|
|
r, g, b = cv2.split(cv2.subtract((*color, 0), image))
|
|
negative = cv2.max(cv2.max(r, g), b)
|
|
return cv2.subtract(255, cv2.add(positive, negative))
|
|
|
|
|
|
def extract_letters(image, letter=(255, 255, 255), threshold=128):
|
|
"""Set letter color to black, set background color to white.
|
|
|
|
Args:
|
|
image: Shape (height, width, channel)
|
|
letter (tuple): Letter RGB.
|
|
threshold (int):
|
|
|
|
Returns:
|
|
np.ndarray: Shape (height, width)
|
|
"""
|
|
r, g, b = cv2.split(cv2.subtract(image, (*letter, 0)))
|
|
positive = cv2.max(cv2.max(r, g), b)
|
|
r, g, b = cv2.split(cv2.subtract((*letter, 0), image))
|
|
negative = cv2.max(cv2.max(r, g), b)
|
|
return cv2.multiply(cv2.add(positive, negative), 255.0 / threshold)
|
|
|
|
|
|
def extract_white_letters(image, threshold=128):
|
|
"""Set letter color to black, set background color to white.
|
|
This function will discourage color pixels (Non-gray pixels)
|
|
|
|
Args:
|
|
image: Shape (height, width, channel)
|
|
threshold (int):
|
|
|
|
Returns:
|
|
np.ndarray: Shape (height, width)
|
|
"""
|
|
r, g, b = cv2.split(cv2.subtract((255, 255, 255, 0), image))
|
|
minimum = cv2.min(cv2.min(r, g), b)
|
|
maximum = cv2.max(cv2.max(r, g), b)
|
|
return cv2.multiply(cv2.add(maximum, cv2.subtract(maximum, minimum)), 255.0 / threshold)
|
|
|
|
|
|
def color_mapping(image, max_multiply=2):
|
|
"""
|
|
Mapping color to 0-255.
|
|
Minimum color to 0, maximum color to 255, multiply colors by 2 at max.
|
|
|
|
Args:
|
|
image (np.ndarray):
|
|
max_multiply (int, float):
|
|
|
|
Returns:
|
|
np.ndarray:
|
|
"""
|
|
image = image.astype(float)
|
|
low, high = np.min(image), np.max(image)
|
|
multiply = min(255 / (high - low), max_multiply)
|
|
add = (255 - multiply * (low + high)) / 2
|
|
image = cv2.add(cv2.multiply(image, multiply), add)
|
|
image[image > 255] = 255
|
|
image[image < 0] = 0
|
|
return image.astype(np.uint8)
|
|
|
|
|
|
def image_left_strip(image, threshold, length):
|
|
"""
|
|
In `DAILY:200/200` strip `DAILY:` and leave `200/200`
|
|
|
|
Args:
|
|
image (np.ndarray): (height, width)
|
|
threshold (int):
|
|
0-255
|
|
The first column with brightness lower than this
|
|
will be considered as left edge.
|
|
length (int):
|
|
Strip this length of image after the left edge
|
|
|
|
Returns:
|
|
np.ndarray:
|
|
"""
|
|
brightness = np.mean(image, axis=0)
|
|
match = np.where(brightness < threshold)[0]
|
|
|
|
if len(match):
|
|
left = match[0] + length
|
|
total = image.shape[1]
|
|
if left < total:
|
|
image = image[:, left:]
|
|
return image
|
|
|
|
|
|
def red_overlay_transparency(color1, color2, red=247):
|
|
"""Calculate the transparency of red overlay.
|
|
|
|
Args:
|
|
color1: origin color.
|
|
color2: changed color.
|
|
red(int): red color 0-255. Default to 247.
|
|
|
|
Returns:
|
|
float: 0-1
|
|
"""
|
|
return (color2[0] - color1[0]) / (red - color1[0])
|
|
|
|
|
|
def color_bar_percentage(image, area, prev_color, reverse=False, starter=0, threshold=30):
|
|
"""
|
|
Args:
|
|
image:
|
|
area:
|
|
prev_color:
|
|
reverse: True if bar goes from right to left.
|
|
starter:
|
|
threshold:
|
|
|
|
Returns:
|
|
float: 0 to 1.
|
|
"""
|
|
image = crop(image, area)
|
|
image = image[:, ::-1, :] if reverse else image
|
|
length = image.shape[1]
|
|
prev_index = starter
|
|
|
|
for _ in range(1280):
|
|
bar = color_similarity_2d(image, color=prev_color)
|
|
index = np.where(np.any(bar > 255 - threshold, axis=0))[0]
|
|
if not index.size:
|
|
return prev_index / length
|
|
else:
|
|
index = index[-1]
|
|
if index <= prev_index:
|
|
return index / length
|
|
prev_index = index
|
|
|
|
prev_row = bar[:, prev_index] > 255 - threshold
|
|
if not prev_row.size:
|
|
return prev_index / length
|
|
prev_color = np.mean(image[:, prev_index], axis=0)
|
|
|
|
return 0.
|