OpenCV入门7——OpenCV中的滤波器（包括低通滤波与高通滤波，其中低通滤波用于降噪，而高通滤波用于边缘检测）

图像滤波

卷积核=滤波器

卷积相关概念

锚点

锚点就是卷积核所对应的图像中间的点，比方说是3x3的卷积核，那对应的锚点可以是16

实战图像卷积

低通滤波：低于某个阀值滤波可以通过
高通滤波：高于某个阀值滤波可以通过

Blur an image with a 2d convolution matrix

# -*- coding: utf-8 -*-
import cv2
import numpy as np

img = cv2.imread('E://pic//10.jpg')

kernel = np.ones((5, 5), np.float32) / 25
dst = cv2.filter2D(img, -1, kernel)

cv2.imshow('dst', dst)
cv2.imshow('img', img)

key = cv2.waitKey(0) & 0xff
if key == ord('q'):
    cv2.destroyAllWindows()

方盒滤波与均值滤波

这两种滤波API功能基本一样，一般我们用blur这个

# -*- coding: utf-8 -*-
import cv2
import numpy as np

img = cv2.imread('E://pic//10.jpg')

# kernel = np.ones((5, 5), np.float32) / 25
# dst = cv2.filter2D(img, -1, kernel)
dst = cv2.blur(img, (5, 5))

cv2.imshow('dst', dst)
cv2.imshow('img', img)

key = cv2.waitKey(0) & 0xff
if key == ord('q'):
    cv2.destroyAllWindows()

高斯滤波

# -*- coding: utf-8 -*-
import cv2
import numpy as np

img = cv2.imread('./gaussian.png')

# kernel = np.ones((5, 5), np.float32) / 25
# dst = cv2.filter2D(img, -1, kernel)
dst = cv2.GaussianBlur(img, (5, 5), sigmaX=1)

cv2.imshow('dst', dst)
cv2.imshow('img', img)

key = cv2.waitKey(0) & 0xff
if key == ord('q'):
    cv2.destroyAllWindows()

中值滤波

中值滤波的优点是对胡椒噪音效果明显

# -*- coding: utf-8 -*-
import cv2
import numpy as np

img = cv2.imread('./papper.png')

# kernel = np.ones((5, 5), np.float32) / 25
# dst = cv2.filter2D(img, -1, kernel)
dst = cv2.medianBlur(img, 5)

cv2.imshow('dst', dst)
cv2.imshow('img', img)

key = cv2.waitKey(0) & 0xff
if key == ord('q'):
    cv2.destroyAllWindows()

双边滤波

可以进行美颜

# -*- coding: utf-8 -*-
import cv2
import numpy as np

img = cv2.imread('./bieber.jpeg')

# kernel = np.ones((5, 5), np.float32) / 25
# dst = cv2.filter2D(img, -1, kernel)
# dst = cv2.medianBlur(img, 5)
dst = cv2.bilateralFilter(img, 7, 20, 50)

cv2.imshow('dst', dst)
cv2.imshow('img', img)

key = cv2.waitKey(0) & 0xff
if key == ord('q'):
    cv2.destroyAllWindows()

高通滤波—索贝尔算子

# -*- coding: utf-8 -*-
import cv2
import numpy as np

img = cv2.imread('./chess.png')

# 索贝尔算子y方向边缘
d1 = cv2.Sobel(img, cv2.CV_64F, 1, 0, ksize=5)
# 索贝尔算子x方向边缘
d2 = cv2.Sobel(img, cv2.CV_64F, 0, 1, ksize=5)

# dst = d1 + d2
dst = cv2.add(d1, d2)

cv2.imshow('img', img)
cv2.imshow('d1', d1)
cv2.imshow('d2', d2)
cv2.imshow('dst', dst)

key = cv2.waitKey(0) & 0xff
if key == ord('q'):
    cv2.destroyAllWindows()

高通滤波—沙尔算子

# -*- coding: utf-8 -*-
import cv2
import numpy as np

img = cv2.imread('./chess.png')

# 索贝尔算子y方向边缘
# d1 = cv2.Sobel(img, cv2.CV_64F, 1, 0, ksize=5)

d1 = cv2.Scharr(img, cv2.CV_64F, 1, 0)
# 索贝尔算子x方向边缘
# d2 = cv2.Sobel(img, cv2.CV_64F, 0, 1, ksize=5)

d2 = cv2.Scharr(img, cv2.CV_64F, 0, 1)

# dst = d1 + d2
dst = cv2.add(d1, d2)

cv2.imshow('img', img)
cv2.imshow('d1', d1)
cv2.imshow('d2', d2)
cv2.imshow('dst', dst)

key = cv2.waitKey(0) & 0xff
if key == ord('q'):
    cv2.destroyAllWindows()

高通滤波—拉普拉斯算子

# -*- coding: utf-8 -*-
import cv2
import numpy as np

img = cv2.imread('./chess.png')

# 索贝尔算子y方向边缘
# d1 = cv2.Sobel(img, cv2.CV_64F, 1, 0, ksize=5)
# 沙尔算子
# d1 = cv2.Scharr(img, cv2.CV_64F, 1, 0)
# 索贝尔算子x方向边缘
# d2 = cv2.Sobel(img, cv2.CV_64F, 0, 1, ksize=5)
# 沙尔算子
# d2 = cv2.Scharr(img, cv2.CV_64F, 0, 1)

# dst = d1 + d2
# dst = cv2.add(d1, d2)

# 拉普拉斯
ldst = cv2.Laplacian(img, cv2.CV_64F, ksize=5)

cv2.imshow('img', img)
# cv2.imshow('d1', d1)
# cv2.imshow('d2', d2)
cv2.imshow('dst', ldst)

key = cv2.waitKey(0) & 0xff
if key == ord('q'):
    cv2.destroyAllWindows()

边缘检测Canny

超过最大值肯定是边缘，低于最小值肯定不是边缘，而介于最大值和最小值之间，如果和超出最大值A是连续的，则C也是边缘，而B就不是边缘

# -*- coding: utf-8 -*-
import cv2
import numpy as np

img = cv2.imread('./lena.png')
dst = cv2.Canny(img, 100, 200)

cv2.imshow("img", img)
cv2.imshow("dst", dst)
key = cv2.waitKey(0) & 0xff
if key == ord('q'):
    cv2.destroyAllWindows()

之后我会持续更新，如果喜欢我的文章，请记得一键三连哦，点赞关注收藏，你的每一个赞每一份关注每一次收藏都将是我前进路上的无限动力 ！！！↖(▔▽▔)↗感谢支持！