MyOpenCV-learning- basic image operations

Image reading and basic attribute operations

• RGB: Three channels, you all know it.

Data read-image

• cv2.imread('filepath', flags): read in the image.

  • filepath is the path
  • flags is the flag to read in the picture
    • cv2.IMREAD_COLOR: color image, ignore alpha channel,
    • cv2.IMREAD_GRAYSCALE: grayscale image, 0
    • cv2.IMREAD_UNCHANGED: read in the complete picture, including the alpha channel,

• Alpha channel: more about it later

upper code

import cv2    #opencv读取的格式是BGR（不是RGB格式）
import matplotlib.pyplot as plt
import numpy as np
%matplotlib inline

img = cv2.imread('cat.jpg')    #读取图片信息

img

array([[[184, 196, 200],
[184, 196, 200],
[184, 196, 200],
…,
[201, 203, 204],
[201, 203, 204],
[203, 202, 204]],

[[184, 196, 200],
[184, 196, 200],
[184, 196, 200],
…,
[201, 203, 204],
[201, 203, 204],
[204, 203, 205]],

[[185, 196, 200],
[185, 196, 200],
[185, 196, 200],
…,
[202, 204, 205],
[202, 204, 205],
[204, 203, 205]],

…,

[[203, 205, 205],
[203, 205, 205],
[203, 205, 205],
…,
[226, 225, 227],
[226, 225, 227],
[225, 224, 226]],

[[204, 206, 206],
[204, 206, 206],
[204, 206, 206],
…,
[226, 225, 227],
[226, 225, 227],
[225, 224, 226]],

[[201, 203, 203],
[202, 204, 204],
[202, 204, 204],
…,
[228, 226, 226],
[228, 226, 226],
[229, 226, 228]]], dtype=uint8)

Notice

• dtype value: 8 bytes
• The dimension depends on the number of brackets, as above, there are 3 dimensions. img stores (height, width, depth) in [h,w,c]. That is to say, the first dimension is the height; the second dimension is the width; the third dimension is the innermost square brackets to indicate the depth, which is BGR.

Image display and wait time setting

• Main function: cv2.show()

code example

#图像的显示，也可以创建多个窗口
cv2.imshow ('image',img)    #第一个参数是窗口的名字，其次为图像。

# cv2.imshow('image2',img)    #创建多个窗口

#等待时间，毫秒级，0表示 任意键 终止
cv2.waitKey(10000)

#删除任何我们建立的窗口
cv2.destroyAllWindows()

# cv2.destroyWindow('image')    #删除特定的窗口，在括号内输入想删除的窗口名

• Note that the size of the picture does not change after the window is enlarged. If you want to change it, see the function below.
• Function to adjust image size: cv2.namesWindow()
  • Initially set the function label to cv2.WINDOW_AUTOSIZE
  • The function label is cv2.WINDOW_NORMAL, you can adjust the window size

cv2.namedWindow('image',cv2.WINDOW_NORMAL)
cv2.imshow('image',img)
cv2.waitKey(0)
cv2.destroyAllWindows()

If you want the code to be clean, just set up a function and frame it.

#执行上述三个函数的函数窗口
def cv_show(name,img):
    cv2.imshow(name,img)
    cv2.waitKey(0)
    cv2.destroyAllWindows()

img.shape

(225, 225, 3)

Grayscale

I mentioned the flags of cv2.imread above, then cv2.IMREAD_GRAYSCALE can read grayscale images. Since there is no RGB, it is two-dimensional.

#灰度图的读取做法
img = cv2.imread('cat.jpg',cv2.IMREAD_GRAYSCALE)
img

array([[196, 196, 196, …, 203, 203, 203],
[196, 196, 196, …, 203, 203, 204],
[196, 196, 196, …, 204, 204, 204],
…,
[205, 205, 205, …, 226, 226, 225],
[206, 206, 206, …, 226, 226, 225],
[203, 204, 204, …, 226, 226, 227]], dtype=uint8)

img.shape

(225, 225)

#图像的显示，也可以创建多个窗口
cv2.imshow ('image',img)
#等待时间，毫秒级，0表示任意键终止
cv2.waitKey(10000)
cv2.destroyAllWindows()

other functions

keep

• Notice that a new picture has been added to the file

cv2.imwrite("mycat.png",img)

True

Format

type(img)    #ndarray格式

numpy.ndarray

Number of pixels

img.size    #size：像素点个数

50625

type of data

img.dtype    #数据类型

dtype(‘uint8’)

Using Matplotlib

img = cv2.imread('cat.jpg')
plt.imshow(img,cmap = 'gray',interpolation='bicubic')
plt.xticks([]),plt.yticks([])
plt.show()

• The picture has chromatic aberration, the reason is that the cv2.imread() interface is used to read the picture, and the read in is BGR format and [0~255], so just convert it to RGB format

img_2 = img[:,:,[2,1,0]]
plt.imshow(img_2)

<matplotlib.image.AxesImage at 0x2608f2faef0>

Hey, it's amazing.

Data Reading - Video

Video is composed of images, which become frames, as you must know.

• cv2.VideoCapture can capture the lens and use numbers to control different devices, such as 0, 1

Run it, don't be intimidated by your big face (not talking about myself

import numpy as np
import cv2

cap = cv2.VideoCapture(0)

while(True):
    ret, frame = cap.read()
    
    gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
    
    cv2.imshow('frame',gray)
    if cv2.waitKey(1) & 0xFF == ord('q'):
        break

cap.release()
cv2.destroyAllWindows()

• save video
• FourCC encoding
• Capture video from webcam, rotate each frame horizontally and save it.

# import numpy as np
# import cv2

# cap = cv2.VideoCapture(0)

# # Define the codec and create VideoWriter object
# fourcc = cv2.cv.FOURCC(*'XVID')
# out = cv2.VideoWriter('output.avi',fourcc, 20.0, (640,480))

# while(cap.isOpened()):
#     ret, frame = cap.read()
#     if ret==True:
#         frame = cv2.flip(frame,0)

#         # write the flipped frame
#         out.write(frame)

#         cv2.imshow('frame',frame)
#         if cv2.waitKey(1) & 0xFF == ord('q'):
#             break
#     else:
#         break

# # Release everything if job is finished
# cap.release()
# out.release()
# cv2.destroyAllWindows()

• If it is a video file, just make the path directly

vc = cv2.VideoCapture('test.mp4')

#检查是否打开正确
if vc.isOpened():    #判断能否打开
    open, frame = vc.read()    #读取帧
else:
    open = False

while open:
    ret, frame = vc.read()
    if frame is None:
        break
    if ret == True:
        gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)    #转换为灰度图
        cv2.imshow('result',gray)
        if cv2.waitKey(1) & 0xFF == 27:    #waitkey数值越大，播放速度越慢
            break
vc.release()
cv2.destroyAllWindows()

Capture part of the image data

Part of the picture that can be intercepted by yourself

img = cv2.imread('cat.jpg')
cat = img[0:200,0:20]
cv_show('cat',cat)

b,g,r = cv2.split(img)

b

array([[184, 184, 184, …, 201, 201, 203],
[184, 184, 184, …, 201, 201, 204],
[185, 185, 185, …, 202, 202, 204],
…,
[203, 203, 203, …, 226, 226, 225],
[204, 204, 204, …, 226, 226, 225],
[201, 202, 202, …, 228, 228, 229]], dtype=uint8)

r.shape

(225, 225)

img = cv2.merge((b,g,r))
img.shape

(225, 225, 3)

Only keep a single channel

• R:G:B correspond to 0:1:2 respectively

• Set the remaining two items to zero

#只保留R通道
cur_img = img.copy()
cur_img[:,:,0] = 0
cur_img[:,:,1] = 0
cv_show('R',cur_img)

#只保留G通道
cur_img = img.copy()
cur_img[:,:,0] = 0
cur_img[:,:,2] = 0
cv_show('G',cur_img)

#只保留B通道
cur_img = img.copy()
cur_img[:,:,1] = 0
cur_img[:,:,2] = 0
cv_show('B',cur_img)

border padding

• Convolution introduction, guide https://mlnotebook.github.io/post/CNN1/ Learn more

Look at the code comments, there are explanations

top_size, bottom_size, left_size, right_size = (50,50,50,50)    #上下左右填充值

#复制法，复制最边缘像素
replicate = cv2.copyMakeBorder(img, top_size, bottom_size, left_size, right_size, borderType = cv2.BORDER_REPLICATE)

#反射法，对感兴趣的图像中的像素在两边进行赋值。例如: fedcba|abcdefgh|hgfedc
reflect = cv2.copyMakeBorder(img, top_size, bottom_size, left_size, right_size, borderType = cv2.BORDER_REFLECT)

#反射法，以最边缘像素为轴，对称。例如：gfedcb|abcdefgh|gfedcba
reflect101 = cv2.copyMakeBorder(img, top_size, bottom_size, left_size, right_size, borderType = cv2.BORDER_REFLECT_101)

#外包装法。例如：bcdefh|abcdefgh|abcdefg
wrap = cv2.copyMakeBorder(img, top_size, bottom_size, left_size, right_size, borderType = cv2.BORDER_WRAP)

#常量法，常数值填充
constant = cv2.copyMakeBorder(img, top_size, bottom_size, left_size, right_size, borderType = cv2.BORDER_CONSTANT, value = 700)#需设置value值，选择常数填充

import matplotlib.pyplot as plt
plt.subplot(231), plt.imshow(img, 'gray'), plt.title('ORIGINAL')
plt.subplot(232), plt.imshow(replicate, 'gray'), plt.title('REPLICATE')
plt.subplot(233), plt.imshow(reflect, 'gray'), plt.title('REFLECT')
plt.subplot(234), plt.imshow(reflect101, 'gray'), plt.title('REFLECT_101')
plt.subplot(235), plt.imshow(wrap, 'gray'), plt.title('WRAP')
plt.subplot(236), plt.imshow(constant, 'gray'), plt.title('CONSTANT')

plt.show()

Numeral Calculations

img_cat = cv2.imread('cat.jpg')
img_dog = cv2.imread('dog.jpg')

• additive constant

img_cat2 = img_cat + 10    #所有像素值加10
img_cat[:5,:,0]

array([[184, 184, 184, …, 201, 201, 203],
[184, 184, 184, …, 201, 201, 204],
[185, 185, 185, …, 202, 202, 204],
[185, 185, 185, …, 202, 202, 205],
[185, 185, 185, …, 203, 203, 206]], dtype=uint8)

img_cat2[:5,:,0]

array([[194, 194, 194, …, 211, 211, 213],
[194, 194, 194, …, 211, 211, 214],
[195, 195, 195, …, 212, 212, 214],
[195, 195, 195, …, 212, 212, 215],
[195, 195, 195, …, 213, 213, 216]], dtype=uint8)

• add two pictures

(img_cat + img_cat2)[:5,:,0]    # 和 对256取余（像素值范围为0~255）

array([[122, 122, 122, …, 156, 156, 160],
[122, 122, 122, …, 156, 156, 162],
[124, 124, 124, …, 158, 158, 162],
[124, 124, 124, …, 158, 158, 164],
[124, 124, 124, …, 160, 160, 166]], dtype=uint8)

• Call the cv2.add function to add the two pictures, but it will cross the boundary, and finally take the maximum value, pay attention

cv2.add(img_cat,img_cat2)[:5,:,0]    # add()函数，越界 取最大值

array([[255, 255, 255, …, 255, 255, 255],
[255, 255, 255, …, 255, 255, 255],
[255, 255, 255, …, 255, 255, 255],
[255, 255, 255, …, 255, 255, 255],
[255, 255, 255, …, 255, 255, 255]], dtype=uint8)

image fusion

• Pay attention to the shape value, the difference cannot be added

img_cat + img_dog    #不可操作，两图像shape值不同，无法相加

---

ValueError Traceback (most recent call last)

in
----> 1 img_cat + img_dog #Not operable, the two images have different shape values ​​and cannot be added

ValueError: operands could not be broadcast together with shapes (225,225,3) (678,1024,3)

img_cat.shape

(225, 225, 3)

• resize() function, change the shape value

img_dog = cv2.resize(img_dog, (225,225))   
img_dog.shape

(225, 225, 3)

blend it

res = cv2.addWeighted(img_cat, 0.7, img_dog, 0.3, 2)

plt.imshow(res)

<matplotlib.image.AxesImage at 0x20032e7fc18>

It's pretty pretty, isn't it~

• cv2.resize() can adjust the size of the image

res = cv2.resize(img,(0,0),fx = 1, fy = 3)
plt.imshow(res)

<matplotlib.image.AxesImage at 0x20032d1f518>

res = cv2.resize(img,(0,0),fx = 3, fy = 1)
plt.imshow(res)

<matplotlib.image.AxesImage at 0x20032e1bcc0>

That's it, there may be additions, welcome to discuss.