def robert(img):
img1 = rgbToGray(img)
img2 = rgbToGray(img)
# img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
r, c = img1.shape
# r_sunnzi = [[-1, -1], [1, 1]]
r_sunnzi_x = [[-1, 0], [0, 1]]
r_sunnzi_y = [[0, -1], [1, 0]]
for x in range(r):
for y in range(c):
if (y + 2 <= c) and (x + 2 <= r):
imgChild = img1[x:x + 2, y:y + 2]
list_robert = r_sunnzi_x * imgChild
img1[x, y] = abs(list_robert.sum()) # 求和加绝对值
for x in range(r):
for y in range(c):
if (y + 2 <= c) and (x + 2 <= r):
imgChild = img2[x:x + 2, y:y + 2]
list_robert = r_sunnzi_y * imgChild
img2[x, y] = abs(list_robert.sum()) # 求和加绝对值
# 转uint8
absX = convertScaleAbs(img1)
absY = convertScaleAbs(img2)
result = addWeighted(absX, 0.5, absY, 0.5, 0)
# result = cv2.addWeighted(img1, 0.5, img2, 0.5, 0)
# cv2.imshow("result",result)
# cv2.waitKey(0)
return result
def idealFilter(img, r, kind):
'''
理想滤波器
:param img: 输入图像
:param r: 滤波器半径
:param kind: 滤波器类型
:return: 滤波后的图像
'''
img = cv2.cvtColor(img, cv2.COLOR_RGBA2BGR) # 四维转三维
img = rgbToGray(img) # 灰度化
f = np.fft.fft2(img) # 傅里叶变换
fshift = np.fft.fftshift(f) # 将低频部分移到中心
# 取绝对值：将复数变化成实数
# 取对数的目的为了将数据变化到0-255
s1 = np.log(np.abs(fshift))
# d1 = make_transform_matrix(r, fshift, s1)
if kind == 0: # 理想低通滤波
d1 = make_transform_matrix(r, fshift, s1)
elif kind == 3: # 理想高通滤波
d1 = 1 - make_transform_matrix(r, fshift, s1)
img_d1 = np.abs(np.fft.ifft2(np.fft.ifftshift(fshift * d1)))
img_d1 = img_d1 / img_d1.max()
img_d1 = img_d1 * 255
img_d1 = img_d1.astype(np.uint8)
return img_d1
def __call__(self, img):
if self._kind == 0:
img = rgbToGray(img)
elif self._kind == 1:
img = image_reverse(img)
elif self._kind == 2:
img = binarization(img)
elif self._kind == 3:
img = gammaTranform(self._c_value, self._γ_value, img)
return img
def GaussianFilter(image, d, kind):
image = cv2.cvtColor(image, cv2.COLOR_RGBA2BGR)
image = rgbToGray(image)
f = np.fft.fft2(image)
fshift = np.fft.fftshift(f)
s1 = np.log(np.abs(fshift))
if kind == 2:
d_matrix = make_transform_matrix(d, image, s1)
elif kind == 5:
d_matrix = 1 - make_transform_matrix(d, image, s1)
img_d1 = np.abs(np.fft.ifft2(np.fft.ifftshift(fshift * d_matrix)))
img_d1 = img_d1 / img_d1.max()
img_d1 = img_d1 * 255
img_d1 = img_d1.astype(np.uint8)
return img_d1
def butterworthSelectFilter(image, d, n, W, kind):
image = cv2.cvtColor(image, cv2.COLOR_RGBA2BGR)
image = rgbToGray(image) # 图像灰度化
f = np.fft.fft2(image) # 图像的傅里叶变换
fshift = np.fft.fftshift(f) # 将低频移动到中心
s1 = np.log(np.abs(fshift))
if kind == 1: # 巴特沃斯带阻滤波器
d_matrix = ButterworthBand(image, W, d, n)
elif kind == 4: # 巴特沃斯带通滤波器
d_matrix = 1 - ButterworthBand(image, W, d, n)
# 与模板相乘后再傅里叶逆变换
img_d1 = np.abs(np.fft.ifft2(np.fft.ifftshift(fshift * d_matrix)))
img_d1 = img_d1 / img_d1.max()
img_d1 = img_d1 * 255
img_d1 = img_d1.astype(np.uint8)
return img_d1
def GaussianSelectFilter(image, d, W, kind):
image = cv2.cvtColor(image, cv2.COLOR_RGBA2BGR)
image = rgbToGray(image)
f = np.fft.fft2(image)
fshift = np.fft.fftshift(f)
s1 = np.log(np.abs(fshift))
if kind == 2: # 高斯带阻滤波器
# d_matrix = make_select_matrix(d,image,s1,W)
d_matrix = GaussianBand(image, W, d)
elif kind == 5: # 高斯带通滤波器
# d_matrix = 1-make_select_matrix(d,image,s1,W)
d_matrix = 1 - GaussianBand(image, W, d)
img_d1 = np.abs(np.fft.ifft2(np.fft.ifftshift(fshift * d_matrix)))
img_d1 = img_d1 / img_d1.max()
img_d1 = img_d1 * 255
img_d1 = img_d1.astype(np.uint8)
return img_d1
def laplacian_filter(img, K_size=3):
H, W, C = img.shape
gray = rgbToGray(img)
# zero padding
pad = K_size // 2
out = zeros((H + pad * 2, W + pad * 2), dtype=float)
out[pad:pad + H, pad:pad + W] = gray.copy().astype(float)
tmp = out.copy()
# laplacian kernle
K = [[0., 1., 0.], [1., -4., 1.], [0., 1., 0.]]
# filtering
for y in range(H):
for x in range(W):
out[pad + y, pad + x] = sum(K * (tmp[y:y + K_size, x:x + K_size]))
out = clip(out, 0, 255)
out = out[pad:pad + H, pad:pad + W].astype(uint8)
return out
def idealSelectFilter(img, r, W, kind):
img = cv2.cvtColor(img, cv2.COLOR_RGBA2BGR)
img = rgbToGray(img)
# 傅里叶变换
f = np.fft.fft2(img)
# 将低频部分移到中心
fshift = np.fft.fftshift(f)
# 取绝对值：将复数变化成实数
# 取对数的目的为了将数据变化到0-255
s1 = np.log(np.abs(fshift))
if kind == 0: # 理想带阻滤波器
d1 = make_select_matrix(r, fshift, s1, W)
elif kind == 3: # 理想带通滤波器
d1 = 1 - make_select_matrix(r, fshift, s1, W)
# 与模板相乘后再傅里叶逆变换
img_d1 = np.abs(np.fft.ifft2(np.fft.ifftshift(fshift * d1)))
img_d1 = img_d1 / img_d1.max()
img_d1 = img_d1 * 255
img_d1 = img_d1.astype(np.uint8)
return img_d1
def idealNotchFilter(img, r, kind):
img = cv2.cvtColor(img, cv2.COLOR_RGBA2BGR)
img = rgbToGray(img)
# 傅里叶变换
f = np.fft.fft2(img)
# 将低频部分移到中心
fshift = np.fft.fftshift(f)
# fshift = fshift.astype(np.uint8)
# 取绝对值：将复数变化成实数
# 取对数的目的为了将数据变化到0-255
s1 = np.log(np.abs(fshift))
if kind == 6:
d1 = make_NotchFilter_matrix(r, fshift, s1)
elif kind == 7:
d1 = 1 - make_NotchFilter_matrix(r, fshift, s1)
img_d1 = np.abs(np.fft.ifft2(np.fft.ifftshift(fshift * d1)))
img_d1 = img_d1 / img_d1.max()
img_d1 = img_d1 * 255
img_d1 = img_d1.astype(np.uint8)
return img_d1
def pseudoColorTrans(img, H, S, V, type):
if img.shape == 4:
img = cv2.cvtColor(img, cv2.COLOR_RGBA2BGR)
img_gray = rgbToGray(img)
if type == 0:
img_color = cv2.applyColorMap(img_gray, cv2.COLORMAP_AUTUMN)
elif type == 1:
img_color = cv2.applyColorMap(img_gray, cv2.COLORMAP_BONE)
elif type == 2:
img_color = cv2.applyColorMap(img_gray, cv2.COLORMAP_JET)
elif type == 3:
img_color = cv2.applyColorMap(img_gray, cv2.COLORMAP_WINTER)
elif type == 4:
img_color = cv2.applyColorMap(img_gray, cv2.COLORMAP_RAINBOW)
elif type == 5:
img_color = cv2.applyColorMap(img_gray, cv2.COLORMAP_OCEAN)
elif type == 6:
img_color = cv2.applyColorMap(img_gray, cv2.COLORMAP_SUMMER)
elif type == 7:
img_color = cv2.applyColorMap(img_gray, cv2.COLORMAP_SPRING)
elif type == 8:
img_color = cv2.applyColorMap(img_gray, cv2.COLORMAP_COOL)
elif type == 9:
img_color = cv2.applyColorMap(img_gray, cv2.COLORMAP_HSV)
elif type == 10:
img_color = cv2.applyColorMap(img_gray, cv2.COLORMAP_PINK)
elif type == 11:
img_color = cv2.applyColorMap(img_gray, cv2.COLORMAP_HOT)
img_color = hsvProcess(img_color, H, S, V)
return img_color
# img_gray = cv2.imread("../pic/beach.png",cv2.IMREAD_GRAYSCALE)
# img_color = cv2.applyColorMap(img_gray,cv2.COLORMAP_JET)
# img = cv2.imread('../pic/beach.png')
# img_gray = pseudoColorTrans(img,type)
# cv2.imshow('img_color',img_gray)
# cv2.waitKey(0)
# cv2.imshow('img_color',img_color)
# cv2.waitKey(0)
def butterworthFilter(image, d, n, kind):
'''
巴特沃斯低通滤波器
:param image: 输入图像
:param d: 滤波半径
:param n: 阶数
:return:
'''
image = cv2.cvtColor(image, cv2.COLOR_RGBA2BGR)
image = rgbToGray(image)
f = np.fft.fft2(image)
fshift = np.fft.fftshift(f)
s1 = np.log(np.abs(fshift))
if kind == 1:
d_matrix = make_transform_matrix(image, d, s1, n)
elif kind == 4:
d_matrix = 1 - make_transform_matrix(image, d, s1, n)
img_d1 = np.abs(np.fft.ifft2(np.fft.ifftshift(fshift * d_matrix)))
# 高通滤波
# img_d1 = np.abs(np.fft.ifft2(np.fft.ifftshift(fshift * (1-d_matrix))))
img_d1 = img_d1 / img_d1.max()
img_d1 = img_d1 * 255
img_d1 = img_d1.astype(np.uint8)
return img_d1
def prewitt_filter(img, K_size=3):
gray = rgbToGray(img)
if len(img.shape) == 3:
H, W, C = img.shape
else:
img = expand_dims(img, axis=-1)
H, W, C = img.shape
# 填充0
pad = K_size // 2
out = zeros((H + pad * 2, W + pad * 2), dtype=float)
out[pad:pad + H, pad:pad + W] = gray.copy().astype(float)
tmp = out.copy()
out_v = out.copy()
out_h = out.copy()
## prewitt 水平方向的核
Kv = [[-1., -1., -1.], [0., 0., 0.], [1., 1., 1.]]
## prewitt 竖直方向的核
Kh = [[-1., 0., 1.], [-1., 0., 1.], [-1., 0., 1.]]
# filtering
for y in range(H):
for x in range(W):
out_v[pad + y,
pad + x] = sum(Kv * (tmp[y:y + K_size, x:x + K_size]))
out_h[pad + y,
pad + x] = sum(Kh * (tmp[y:y + K_size, x:x + K_size]))
out_v = clip(out_v, 0, 255)
out_h = clip(out_h, 0, 255)
out_v = out_v[pad:pad + H, pad:pad + W].astype(uint8)
out_h = out_h[pad:pad + H, pad:pad + W].astype(uint8)
dst = addWeighted(out_v, 0.5, out_h, 0.5, 0)
return dst
# 读取图像
# img = cv2.imread('../pic/boy.png')
# prewitt_filter(img)