OpenCV高级图像处理学习笔记

4.1 混合两张图像

#Import required packages
import cv2

#Read image 1
img1=cv2.imread("C:\\Users\\mac\\Pictures\\measure\\cat1.jpg")
#Read image 2
img2=cv2.imread("C:\\Users\\mac\\Pictures\\measure\\cat2.jpg")

#Define alpha and beta
alpha=0.30
beta=0.70

#Blend images
final_image=cv2.addWeighted(img1,alpha,img2,beta,0.0)

#show image
cv2.imshow('mix',final_image)
cv2.waitKey(0)
cv2.destroyAllWindows()

将两张猫咪的图片混合了起来。

4.2 改变图像的对比度和亮度

#Import required packages
import cv2
import numpy as np

#Read images
image=cv2.imread("C:\\Users\\mac\\Pictures\\measure\\cat1.jpg")

#Create a dummy image that stores different contrast and brightness
new_image=np.zeros(image.shape,image.dtype)

#Brightness and contrast parameters
contrast=3.0
bright=2

#Change the contrast and brightness  
for y in range(image.shape[0]):
    for x in range(image.shape[1]):
        for c in range(image.shape[2]):
            new_image[y,x,c]=np.clip(contrast*image[y,x,c]+bright,0,255)
cv2.namedWindow('original', cv2.WINDOW_NORMAL)
cv2.imshow('original',image)
cv2.namedWindow('transform', cv2.WINDOW_NORMAL)
cv2.imshow('transform',new_image)
cv2.waitKey(0)
cv2.destroyAllWindows()

4.3 往图像中添加文字

cv2模块中的cv2.putText()函数可以往图像中添加文字。

#import required packages
import cv2

#Read image
image=cv2.imread("C:\\Users\\mac\\Pictures\\measure\\cat1.jpg")

#Define font
font=cv2.FONT_HERSHEY_SIMPLEX

#Write on the image
cv2.putText(image,"I am a Cat",(230,50),font,0.8,(0,255,0),2,cv2.LINE_AA)
cv2.namedWindow('AddText', cv2.WINDOW_NORMAL)
cv2.imshow('AddText',image)
cv2.waitKey(0)
cv2.destroyAllWindows()

4.4 平滑图像

三种用于平滑图像的滤波器

• 中值滤波器
• 高斯滤波器
• 双边滤波器

4.4.1 中值滤波器

cv2.medianBlur()函数能够实现中值模糊的效果。

4.4.2 高斯滤波器

cv2.GaussianBlur()函数应用于高斯滤波器。

4.4.3 双边滤波器

cv2.bilateralFilter()函数

#Import required packages
import cv2

#Read images for different blurring purposes
image_Original=cv2.imread("C:\\Users\\mac\\Pictures\\measure\\cat1.jpg")
image_MedianBlur=cv2.imread("C:\\Users\\mac\\Pictures\\measure\\cat1.jpg")
image_GaussianBlur=cv2.imread("C:\\Users\\mac\\Pictures\\measure\\cat1.jpg")
image_BilateralBlur=cv2.imread("C:\\Users\\mac\\Pictures\\measure\\cat1.jpg")

#Blur images
image_MedianBlur=cv2.medianBlur(image_MedianBlur,9)
image_GaussianBlur=cv2.GaussianBlur(image_GaussianBlur,(9,9),10)
image_BilateralBlur=cv2.bilateralFilter(image_BilateralBlur,9,100,75)

#show images
cv2.namedWindow('image_MedianBlur', cv2.WINDOW_NORMAL)
cv2.imshow('image_MedianBlur',image_MedianBlur)
cv2.namedWindow('image_GaussianBlur', cv2.WINDOW_NORMAL)
cv2.imshow('image_GaussianBlur',image_GaussianBlur)
cv2.namedWindow('image_BilateralBlur', cv2.WINDOW_NORMAL)
cv2.imshow('image_BilateralBlur',image_BilateralBlur)
cv2.waitKey(0)
cv2.destroyAllWindows()

4.5 改变图像的形状

两种改变图像的操作：侵蚀（erosion）和扩张（dilation）。扩张操作导致物体边界的像素增加，侵蚀操作导致物体边界的像素减少。
cv2.getStructuringElement()函数用于定义核，作为侵蚀或扩张函数的参数。

# EROSION CODE:

#Import package
import cv2

#Read image
image=cv2.imread("C:\\Users\\mac\\Pictures\\measure\\cat1.jpg")

#Define erosion size
s1=0
s2=10
s3=10

#Define erosion type
t1=cv2.MORPH_RECT
t2=cv2.MORPH_CROSS
t3=cv2.MORPH_ELLIPSE

#Define and save the erosion template
tmp1=cv2.getStructuringElement(t1,(2*s1+1,2*s1+1),(s1,s1))
tmp2=cv2.getStructuringElement(t2,(2*s2+1,2*s2+1),(s2,s2))
tmp3=cv2.getStructuringElement(t3,(2*s3+1,2*s3+1),(s3,s3))

#Apply the erosion template to the image and save in different variables
final1=cv2.erode(image,tmp1)
final2=cv2.erode(image,tmp2)
final3=cv2.erode(image,tmp3)

#Show all the images with different erosions
cv2.namedWindow('final1', cv2.WINDOW_NORMAL)
cv2.imshow('final1',final1)
cv2.namedWindow('final2', cv2.WINDOW_NORMAL)
cv2.imshow('final2',final2)
cv2.namedWindow('final3', cv2.WINDOW_NORMAL)
cv2.imshow('final3',final3)
cv2.waitKey(0)
cv2.destroyAllWindows()

# DILATION CODE:

#Import package
import cv2

#Read image
image=cv2.imread("C:\\Users\\mac\\Pictures\\measure\\cat1.jpg")

#Define erosion size
d1=0
d2=10
d3=20

#Define erosion type
t1=cv2.MORPH_RECT
t2=cv2.MORPH_CROSS
t3=cv2.MORPH_ELLIPSE

#Define and save the erosion template
tmp1=cv2.getStructuringElement(t1,(2*d1+1,2*d1+1),(d1,d1))
tmp2=cv2.getStructuringElement(t2,(2*d2+1,2*d2+1),(d2,d2))
tmp3=cv2.getStructuringElement(t3,(2*d3+1,2*d3+1),(d3,d3))

#Apply the erosion template to the image and save in different variables
final1=cv2.dilate(image,tmp1)
final2=cv2.dilate(image,tmp2)
final3=cv2.dilate(image,tmp3)

#Show all the images with different erosions
cv2.namedWindow('final1', cv2.WINDOW_NORMAL)
cv2.imshow('final1',final1)
cv2.namedWindow('final2', cv2.WINDOW_NORMAL)
cv2.imshow('final2',final2)
cv2.namedWindow('final3', cv2.WINDOW_NORMAL)
cv2.imshow('final3',final3)
cv2.waitKey(0)
cv2.destroyAllWindows()

4.6 实现图像阈值化

首先需要将图像转换成灰度格式，然后转换成二值格式——只有黑色与白色的图像。

提供一个参考值，然后将所有值大于或小于它的像素都转换成黑色或白色。
使用cv2.threshold()函数做图像阈限化。

#Import packages
import cv2

#Read image
image=cv2.imread("C:\\Users\\mac\\Pictures\\measure\\cat1.jpg")

#Define threshold types
'''
0 - Binary
1 - Binary Inverted
2 - Truncated
3 - Threshold To Zero
4 - Threshold To Zero Inverted
'''

#Apply different threshold and save in different variables
_, img1=cv2.threshold(image,50,255,0)
_, img2=cv2.threshold(image,50,255,1)
_, img3=cv2.threshold(image,50,255,2)
_, img4=cv2.threshold(image,50,255,3)
_, img5=cv2.threshold(image,50,255,4)

#Show the different threshold images
cv2.namedWindow('img1', cv2.WINDOW_NORMAL)
cv2.imshow('img1',img1)
cv2.namedWindow('img2', cv2.WINDOW_NORMAL)
cv2.imshow('img2',img2)
cv2.namedWindow('img3', cv2.WINDOW_NORMAL)
cv2.imshow('img3',img3)
cv2.namedWindow('img4', cv2.WINDOW_NORMAL)
cv2.imshow('img4',img4)
cv2.namedWindow('img5', cv2.WINDOW_NORMAL)
cv2.imshow('img5',img5)
cv2.waitKey(0)
cv2.destroyAllWindows()

4.7 计算梯度

在本节中，将介绍如何使用索伯导数（Sobel derivative）做边缘检测。边有两种方向：垂直方向和水平方向。。针对这种算法，我们强调只有空间频率非常高的区域才能算作边。一个区域的空间频率是指这个区域内的细节丰富程度。

在下面的代码中，我们用高斯模糊去除噪声，然后将图像转换成灰度图。我们使用cv2.cvtColor()函数将图像转换成灰度图。skimage模块里的函数也能达到同样的效果。最后，我们将灰度图输出到cv2.Sobel()函数。