def mission2():
img1 = cv2.imread(SAVEDPATH + "\guass_noise.jpg", cv2.IMREAD_GRAYSCALE)
img2 = cv2.imread(SAVEDPATH + "\salt_and_pepper_noise.jpg",
cv2.IMREAD_GRAYSCALE)
show(img1)
show(img2)
median_filter(img1)
mean_filter(img2)
mean_filter(img1)
mean_filter(img2)
def homomorphic(img:np.ndarray):
img = img + 0.000000001
print(img)
img = np.log(img)
print(img)
img = np.fft.fft(img)
img = BHPF(img)
img = np.fft.ifft(img)
img = np.exp(img)
print(img)
img = np.array(img, dtype='f')
print(img)
show(img)
return img
def gauss_noise(picture: np.ndarray,
SNR: float = 0.75,
verbose: bool = True) -> np.ndarray:
orig_size = picture.shape
arr = picture.flatten()
length = arr.shape[0]
choice = np.random.randint(0, length / 3, int((length / 3) * (1 - SNR)))
for i in choice:
x = np.random.random()
arr[i * 3] = np.random.normal(arr[i * 3], 20)
arr[i * 3 + 1] = np.random.normal(arr[i * 3 + 1], 20)
arr[i * 3 + 2] = np.random.normal(arr[i * 3 + 2], 20)
picture = arr.reshape(orig_size)
if verbose:
show(picture)
cv2.imwrite(SAVEDPATH + r'\guass_noise.jpg', picture)
return picture
#sppicture = salt_and_pepper_noise(picture)
#gspicture = gauss_noise(picture)
def salt_and_pepper_noise(picture: np.ndarray,
SNR: float = 0.75,
verbose: bool = True) -> np.ndarray:
orig_size = picture.shape
arr = picture.flatten()
length = arr.shape[0]
choice = np.random.randint(0, length / 3, int((length / 3) * (1 - SNR)))
for i in choice:
x = np.random.random()
if x < 0.5:
arr[i * 3] = 0
arr[i * 3 + 1] = 0
arr[i * 3 + 2] = 0
else:
arr[i * 3] = 255
arr[i * 3 + 1] = 255
arr[i * 3 + 2] = 255
picture = arr.reshape(orig_size)
if verbose:
show(picture)
cv2.imwrite(SAVEDPATH + r'\salt_and_pepper_noise.jpg', picture)
return picture
from Script.lab1.noise import gauss_noise
def bilateral(img: np.ndarray, sigmad: float = 200, sigmar: float = 200):
img = np.array(img, dtype=np.float)
g = np.zeros(img.shape, dtype=np.float)
img = samepadding(img, np.eye(5))
for i0 in range(0, img.shape[0] - 4):
for j0 in range(0, img.shape[1] - 4):
i = i0 + 2
j = j0 + 2
sumup = 0
sumdown = 0
for k in range(i0, i0 + 5):
for l in range(j0, j0 + 5):
d = np.exp(-(((i - k)**2 + (j - l)**2) / (2 * sigmad**2)))
r = np.exp(-((img[i][j] - img[k][l])**2) / (2 * sigmar**2))
w = d * r
sumdown += w
sumup += w * img[k][l]
g[i0][j0] = sumup / sumdown
g = np.array(g, dtype='u1')
return g
img = get()
show(img)
img = gauss_noise(img, SNR=0.95)
img1 = bilateral(img)
show(img1)
def test_sobel():
img = get()
img = sobel(img)
show(img)
#test_sobel()
import numpy as np
import cv2
from configuration import IMAGEPATH
from configuration import SAVEDPATH
from Script.util.show import show
picture = cv2.imread(IMAGEPATH + '\lena.jpg')
show(picture)
def salt_and_pepper_noise(picture: np.ndarray,
SNR: float = 0.75,
verbose: bool = True) -> np.ndarray:
orig_size = picture.shape
arr = picture.flatten()
length = arr.shape[0]
choice = np.random.randint(0, length / 3, int((length / 3) * (1 - SNR)))
for i in choice:
x = np.random.random()
if x < 0.5:
arr[i * 3] = 0
arr[i * 3 + 1] = 0
arr[i * 3 + 2] = 0
else:
arr[i * 3] = 255
arr[i * 3 + 1] = 255
arr[i * 3 + 2] = 255
picture = arr.reshape(orig_size)
if verbose:
show(picture)
cv2.imwrite(SAVEDPATH + r'\salt_and_pepper_noise.jpg', picture)
from Script.lab1.getpicture import get
from Script.lab1.noise import gauss_noise
def BHPF(img:np.ndarray,D0=360,n=14):
filter = np.zeros(img.shape)
for i in range(0,filter.shape[0]):
for j in range(0,filter.shape[1]):
filter[i][j] = np.sqrt((i-filter.shape[0]//2)**2 + (j-filter.shape[1]//2)**2)
img1 = img + 0.000000001
return img * (1/(1+(D0/filter)**(2*n)))
def homomorphic(img:np.ndarray):
img = img + 0.000000001
print(img)
img = np.log(img)
print(img)
img = np.fft.fft(img)
img = BHPF(img)
img = np.fft.ifft(img)
img = np.exp(img)
print(img)
img = np.array(img, dtype='f')
print(img)
show(img)
return img
img = get()
img = gauss_noise(img)
show(img)
img = img / 255
homomorphic(img)
ans[i][j - 1] = 255
ans[i][j + 1] = 255
ans[i][j + 2] = 255
return ans
@classmethod
def open(cls, img):
img = cls.erode(img)
img = cls.dilate(img)
return img
@classmethod
def close(cls, img):
img = cls.dilate(img)
img = cls.erode(img)
return img
img = get()
img = morphology.binaryzation(img)
# for i in range(0,10):
# img = morphology.open(img)
# img = morphology.close(img)
er = morphology.erode(img)
di = morphology.dilate(img)
ans = di - er
ans = np.abs(ans)
show(ans)
def getlisenceplate(car: np.ndarray):
bluef = car[..., 0]
bluef = np.array(bluef, dtype='f4')
anotherf = car[..., 1]
anotherf = np.array(anotherf, dtype='f4')
anotherf1 = car[..., 2]
anotherf1 = np.array(anotherf1, dtype='f4')
ansf1 = bluef - anotherf
ansf2 = bluef - anotherf1
ansf = (ansf1 + ansf2) // 2
ansf = (np.abs(ansf) + ansf) // 2
print(ansf)
ansf = np.array(ansf, dtype='u1')
# cv2.imshow('test',ansf)
# cv2.waitKey(0)
# cv2.destroyAllWindows()
#show(ansf)
#ret,ansf = cv2.threshold(ansf,80,255,cv2.THRESH_BINARY)
ansfx = cv2.Sobel(ansf, -1, 1, 0)
ansfy = cv2.Sobel(ansf, -1, 0, 1)
ansf = np.array(ansfx + ansfy, dtype='u1')
kernel = cv2.getStructuringElement(cv2.MORPH_RECT, (3, 3))
for i in range(0, 9):
ansf = cv2.dilate(ansf, kernel)
for i in range(0, 9):
ansf = cv2.erode(ansf, kernel)
for i in range(0, 5):
ansf = cv2.erode(ansf, kernel)
for i in range(0, 5):
ansf = cv2.dilate(ansf, kernel)
ret, ansf = cv2.threshold(ansf, 60, 255, cv2.THRESH_BINARY)
#show(ansf)
# ansfx = cv2.Sobel(ansf, -1, 1, 0)
# ansfy = cv2.Sobel(ansf, -1, 0, 1)
# ansf = np.array(ansfx + ansfy,dtype='u1')
ansf = cv2.morphologyEx(ansf, cv2.MORPH_GRADIENT, kernel)
#ansf = cv2.dilate(ansf, kernel)
#show(ansf)
# lines = cv2.HoughLines(ansf, 0.8, np.pi/180, 40)
# board = np.zeros((ansf.shape[0],ansf.shape[1],3))
# for line in lines:
# rho, theta = line[0]
# a = np.cos(theta)
# b = np.sin(theta)
# x0 = a * rho
# y0 = b * rho
# x1 = int(x0 + 1000 * (-b))
# y1 = int(y0 + 1000 * (a))
# x2 = int(x0 - 1000 * (-b))
# y2 = int(y0 - 1000 * (a))
# cv2.line(board, (x1, y1), (x2, y2), (0, 0, 255))
# show(board)
ret, ansf = cv2.threshold(ansf, 254, 255, cv2.THRESH_BINARY)
#show(ansf)
xmin = ansf.shape[1] - 1
xmax = 0
ymin = ansf.shape[0] - 1
ymax = 0
for i in range(0, ansf.shape[0]):
for j in range(0, ansf.shape[1]):
if ansf[i][j] >= 254:
if xmin > j:
xmin = j
if ymin > i:
ymin = i
if xmax < j:
xmax = j
if ymax < i:
ymax = i
cv2.line(car, (xmin, ymin), (xmax, ymin), (0, 0, 255), 3)
cv2.line(car, (xmin, ymax), (xmax, ymax), (0, 0, 255), 3)
cv2.line(car, (xmin, ymin), (xmin, ymax), (0, 0, 255), 3)
cv2.line(car, (xmax, ymin), (xmax, ymax), (0, 0, 255), 3)
show(car)
return car
def median_filter(img):
img = medianFilter22(img, (3, 3))
show(img)
return img
def mean_filter(img):
core = 1 / 9 * np.ones((3, 3))
img = conv(img, core)
show(img)
return img
