def median_filter(img):
"""九点中值滤波"""
dim = 3
n = dim * dim
mid_pos = int((n + 1) / 2)
margin = int((dim - 1) / 2)
shape = img.shape
new_img = immake(shape) # 构造新像素矩阵
rows, cols = shape
for i in range(0, rows):
for j in range(0, cols):
# 边缘像素不处理
if i <= margin or i >= rows - margin or j <= margin or j >= cols - margin:
new_img[i, j] = img[i, j]
continue
# 邻域像素排序后取中值
arr = [
img[x, y] for x in range(i - margin, i + margin + 1)
for y in range(j - margin, j + margin + 1)
]
arr.sort()
new_img[i, j] = arr[mid_pos]
return new_img
def linear_filter(img):
"""3*3九点均值滤波"""
operator = [[1, 1, 1], [1, 1, 1], [1, 1, 1]]
size = 9
dim = len(operator)
shape = img.shape
new_img = immake(shape) # 构造新像素矩阵
rows, cols = shape
for i in range(rows):
for j in range(cols):
r = 0
for m in range(dim):
# 超出边界行的元素按边界值处理
x = max(i + m - 1, 0)
x = min(x, rows - 1)
for n in range(dim):
# 超出边界列的元素按边界值处理
y = max(j + n - 1, 0)
y = min(y, cols - 1)
# 与相应模板值做卷积
w = operator[m][n]
r += w * img[x, y]
new_img[i, j] = r / size
return new_img
def roberts_sharpen(img):
"""Roberts交叉差分锐化滤波"""
shape = img.shape
new_img = immake(shape)
rows, cols = shape
for i in range(0, rows - 1):
for j in range(0, cols - 1):
grad = abs(int(img[i, j]-int(img[i+1, j+1]))) + \
abs(int(img[i+1, j])-int(img[i, j+1]))
new_img[i, j] = grad
return new_img
def grad_sharpen(img):
"""梯度法锐化滤波"""
shape = img.shape
new_img = immake(shape)
rows, cols = shape
for i in range(0, rows):
for j in range(0, cols):
grad = 0
if i < rows - 1:
grad += abs(int(img[i, j]) - int(img[i + 1, j]))
if j < cols - 1:
grad += abs(int(img[i, j]) - int(img[i, j + 1]))
new_img[i, j] = grad
return new_img
def laplacian_enhance(img):
operator = [
[1, 1, 1],
[1, -9, 1],
[1, 1, 1],
]
shape = img.shape
new_img = immake(shape)
rows, cols = shape
for i in range(1, rows - 1):
for j in range(1, cols - 1):
s = 0
for m in range(0, 3):
for n in range(0, 3):
w = operator[m][n]
x = i - 1 + m
y = j - 1 + n
s += w * img[x][y]
new_img[i, j] = abs(int(s))
return new_img
def laplacian_sharpen(img):
"""3*3中心点为-8的掩模拉普拉斯锐化滤波"""
operator = [
[1, 1, 1],
[1, -8, 1],
[1, 1, 1],
]
shape = img.shape
new_img = immake(shape)
rows, cols = shape
for i in range(1, rows - 1):
for j in range(1, cols - 1):
s = 0
for m in range(0, 3):
for n in range(0, 3):
w = operator[m][n]
x = i - 1 + m
y = j - 1 + n
s += w * img[x][y]
new_img[i, j] = abs(int(s))
return new_img