class NoiseGenerator(object):
def generated_noise(self, img, num):
print("*****************")
for i1, char in enumerate(num):
if num[i1] == 0:
img = self.original(img)
print("原图")
elif num[i1] == 1:
img = self.gaussian_img(img)
print("添加了高斯模糊")
elif num[i1] == 2:
img = self.salt_img(img)
print("添加了椒盐噪声")
elif num[i1] == 3:
img = self.liner(img, a=2)
print("增加了对比度")
elif num[i1] == 4:
img = self.liner(img, a=0.5)
print("降低了对比度")
return img
def gaussian_img(self, img): # 高斯模糊
im = cv2.GaussianBlur(img, (9, 9), 5)
return im
def salt_img(self, img): # 椒盐噪声
n = int(img.shape[0] * img.shape[1] * 0.1)
ilist = np.random.randint(0, img.shape[1], n)
jlist = np.random.randint(0, img.shape[0], n)
for k in range(n):
i = ilist[k]
j = jlist[k]
if img.ndim == 2:
img[j, i] = 255
elif img.ndim == 3:
img[j:j + 1, i:i + 1, :] = 255
return img
def liner(self, img, a): # 线性变换，增强图像的对比度
out_image = float(a) * img
# 进行数据截断, 大于255的值要截断为255
out_image[out_image > 255] = 255
# 数据类型转化
out_image = np.round(out_image)
out_image = out_image.astype(np.uint8)
return out_image
def original(self, img):
return img
if __name__ == '__main__':
rd = NoiseGenerator() # 调用创建的随机噪声类
filenames = './positive_sample' # 图像所在文件夹
a = list(range(5)) # 最多2种噪声的叠加, range(A): 0 ~ A-1
b = []
# 随机抽取两类噪声所有的类型，这里相当于在五个数中随机抽取两个数
for i in itertools.product(a, repeat=2):
b.append(i)
for index, char in enumerate(b):
# 获取所有可能类型的索引和数值，如index=0, char="0.jpg"
i_name = 0 # 保存图像名称
for image_name in os.listdir(filenames): # image_name = "0.jpg"
c = b[index] # c=char，为了写循环，将char获取的图像名称用取索引的方式得到
img_path = filenames + '/' + image_name # 图像路径
img = cv2.imread(img_path, cv2.IMREAD_GRAYSCALE) # 读取图像
# 随机产生噪声，例如c=[2,3],则添加椒盐噪声和增加图像对比度
result_image = rd.generated_noise(img, c)
# cv2.imshow("img1", result_image)
# cv2.waitKey(0)
index += 1
# 在循环一次b中的所有情况后，index取得最大值25，将indedx置为0，继续循环
if index == 25:
index = 0
cv2.imwrite('./add_noise_sample/{}.jpg'.format(i_name), result_image)
i_name += 1
|