首先，為什么我們需要裁剪？進行裁剪以從圖像中刪除所有不需要的對象或區(qū)域。甚至突出圖像的特定特征。

以下代碼片段顯示了如何使用 Python 和 C++ 裁剪圖像。在這篇文章中，您將詳細了解這些內(nèi)容。

Python

1	# Import packages

2	import cv2

3	import numpy as np

4

5	img = cv2.imread('test.jpg')

6	print(img.shape) # Print image shape

7	cv2.imshow("original", img)

8

9	# Cropping an image

10	cropped_image = img[80:280, 150:330]

11

12	# Display cropped image

13	cv2.imshow("cropped", cropped_image)

14

15	# Save the cropped image

16	cv2.imwrite("Cropped Image.jpg", cropped_image)

17

18	cv2.waitKey(0)

19	cv2.destroyAllWindows()

C++

1	// Include Libraries

2	#include<opencv2/opencv.hpp>

3	#include<iostream>

4

5	// Namespace nullifies the use of cv::function();

6	using namespace std;

7	using namespace cv;

8

9	int main()

10

{

11	// Read image

12	Mat img = imread("test.jpg");

13	cout << "Width : " << img.size().width << endl;

14	cout << "Height: " << img.size().height << endl;

15	cout<<"Channels: :"<< img.channels() << endl;

16	// Crop image

17	Mat cropped_image = img(Range(80,280), Range(150,330));

18

19	//display image

20	imshow(" Original Image", img);

21	imshow("Cropped Image", cropped_image);

22

23	//Save the cropped Image

24	imwrite("Cropped Image.jpg", cropped_image);

25

26	// 0 means loop infinitely

27	waitKey(0);

28	destroyAllWindows();

29

return 0;

30

}

使用 OpenCV 進行裁剪

在這篇文章中將用于裁剪的圖像。

Python：

1	img=cv2.imread('test.png')

2

3	# Prints Dimensions of the image

4	print(img.shape)

5

6	# Display the image

7	cv2.imshow("original", img)

8	cv2.waitKey(0)

9	cv2.destroyAllWindows()

C++

1	Mat img = imread("test.jpg");

2

3	//Print the height and width of the image

4	cout << "Width : " << img.size().width << endl;

5	cout << "Height: " << img.size().height << endl;

6	cout << "Channels: " << img.channels() << endl;

7

8	// Display image

9	imshow("Image", img);

10	waitKey(0);

11	destroyAllWindows();

上面的代碼讀取并顯示圖像及其尺寸。維度不僅包括二維矩陣的寬度和高度，還包括通道數(shù)（例如，RGB 圖像有 3 個通道——紅色、綠色和藍色）。

讓我們嘗試裁剪包含花朵的圖像部分。

Python

1	cropped_image = img[80:280, 150:330] # Slicing to crop the image

2

3	# Display the cropped image

4	cv2.imshow("cropped", cropped_image)

5	cv2.waitKey(0)

6	cv2.destroyAllWindows()

C++

1	Mat crop = img(Range(80,280),Range(150,330)); // Slicing to crop the image

2

3	// Display the cropped image

4	imshow("Cropped Image", crop);

5

6	waitKey(0);

7	destroyAllWindows();

8

return 0;

在 Python 中，您使用與 NumPy 數(shù)組切片相同的方法裁剪圖像。要對數(shù)組進行切片，您需要指定第一維和第二維的開始和結(jié)束索引。 

• 第一個維度始終是圖像的行數(shù)或高度。

• 第二個維度是圖像的列數(shù)或?qū)挾取?nbsp;

二維數(shù)組的第一個維度表示數(shù)組的行（其中每一行表示圖像的 y 坐標），這符合慣例。如何對 NumPy 數(shù)組進行切片？查看此示例中的語法：

cropped = img[start_row:end_row, start_col:end_col]

在 C++ 中，我們使用該Range()函數(shù)來裁剪圖像。 

• 與 Python 一樣，它也適用于切片。 

• 在這里，圖像也被讀取為 2D 矩陣，遵循上述相同的約定。 

以下是裁剪圖像的 C++ 語法：

img(Range(start_row, end_row), Range(start_col, end_col))

使用裁剪將圖像分成小塊

OpenCV 中裁剪的一種實際應(yīng)用是將圖像分割成更小的塊。使用循環(huán)從圖像中裁剪出一個片段。首先從圖像的形狀中獲取所需補丁的高度和寬度。

Python

1	img =  cv2.imread("test_cropped.jpg")

2	image_copy = img.copy()

3	imgheight=img.shape[0]

4	imgwidth=img.shape[1]

C++

1	Mat img = imread("test_cropped.jpg");

2	Mat image_copy = img.clone();

3	int imgheight = img.rows;

4	int imgwidth = img.cols;

加載高度和寬度以指定需要裁剪較小補丁的范圍。為此，請使用range()Python 中的函數(shù)。for現(xiàn)在，使用兩個循環(huán)進行裁剪：

1. 一個用于寬度范圍

2. 其他為高度范圍 

我們使用高度和寬度分別為 76 像素和 104 像素的補丁。內(nèi)部和外部循環(huán)的步幅（我們在圖像中移動的像素數(shù)）等于我們正在考慮的補丁的寬度和高度。

Python

1

M = 76

2

N = 104

3

x1 = 0

4

y1 = 0

5

6	for y in range(0, imgheight, M):

7	for x in range(0, imgwidth, N):

8	if (imgheight - y) < M or (imgwidth - x) < N:

9

break

10

11	y1 = y + M

12	x1 = x + N

13

14	# check whether the patch width or height exceeds the image width or height

15	if x1 >= imgwidth and y1 >= imgheight:

16	x1 = imgwidth - 1

17	y1 = imgheight - 1

18	#Crop into patches of size MxN

19	tiles = image_copy[y:y+M, x:x+N]

20	#Save each patch into file directory

21	cv2.imwrite('saved_patches/'+'tile'+str(x)+'_'+str(y)+'.jpg', tiles)

22	cv2.rectangle(img, (x, y), (x1, y1), (0, 255, 0), 1)

23	elif y1 >= imgheight: # when patch height exceeds the image height

24	y1 = imgheight - 1

25	#Crop into patches of size MxN

26	tiles = image_copy[y:y+M, x:x+N]

27	#Save each patch into file directory

28	cv2.imwrite('saved_patches/'+'tile'+str(x)+'_'+str(y)+'.jpg', tiles)

29	cv2.rectangle(img, (x, y), (x1, y1), (0, 255, 0), 1)

30	elif x1 >= imgwidth: # when patch width exceeds the image width

31	x1 = imgwidth - 1

32	#Crop into patches of size MxN

33	tiles = image_copy[y:y+M, x:x+N]

34	#Save each patch into file directory

35	cv2.imwrite('saved_patches/'+'tile'+str(x)+'_'+str(y)+'.jpg', tiles)

36	cv2.rectangle(img, (x, y), (x1, y1), (0, 255, 0), 1)

37

else:

38	#Crop into patches of size MxN

39	tiles = image_copy[y:y+M, x:x+N]

40	#Save each patch into file directory

41	cv2.imwrite('saved_patches/'+'tile'+str(x)+'_'+str(y)+'.jpg', tiles)

42	cv2.rectangle(img, (x, y), (x1, y1), (0, 255, 0), 1)

C++

1	int M = 76;

2	int N = 104;

3

4	int x1 = 0;

5	int y1 = 0;

6	for (int y = 0; y<imgheight; y=y+M)

7

{

8	for (int x = 0; x<imgwidth; x=x+N)

9

{

10	if ((imgheight - y) < M || (imgwidth - x) < N)

11

{

12

break;

13

}

14	y1 = y + M;

15	x1 = x + N;

16	string a = to_string(x);

17	string b = to_string(y);

18

19	if (x1 >= imgwidth && y1 >= imgheight)

20

{

21	x = imgwidth - 1;

22	y = imgheight - 1;

23	x1 = imgwidth - 1;

24	y1 = imgheight - 1;

25

26	// crop the patches of size MxN

27	Mat tiles = image_copy(Range(y, imgheight), Range(x, imgwidth));

28	//save each patches into file directory

29	imwrite("saved_patches/tile" + a + '_' + b + ".jpg", tiles);

30	rectangle(img, Point(x,y), Point(x1,y1), Scalar(0,255,0), 1);

31

}

32	else if (y1 >= imgheight)

33

{

34	y = imgheight - 1;

35	y1 = imgheight - 1;

36

37	// crop the patches of size MxN

38	Mat tiles = image_copy(Range(y, imgheight), Range(x, x+N));

39	//save each patches into file directory

40	imwrite("saved_patches/tile" + a + '_' + b + ".jpg", tiles);

41	rectangle(img, Point(x,y), Point(x1,y1), Scalar(0,255,0), 1);

42

}

43	else if (x1 >= imgwidth)

44

{

45	x = imgwidth - 1;

46	x1 = imgwidth - 1;

47

48	// crop the patches of size MxN

49	Mat tiles = image_copy(Range(y, y+M), Range(x, imgwidth));

50	//save each patches into file directory

51	imwrite("saved_patches/tile" + a + '_' + b + ".jpg", tiles);

52	rectangle(img, Point(x,y), Point(x1,y1), Scalar(0,255,0), 1);

53

}

54

else

55

{

56	// crop the patches of size MxN

57	Mat tiles = image_copy(Range(y, y+M), Range(x, x+N));

58	//save each patches into file directory

59	imwrite("saved_patches/tile" + a + '_' + b + ".jpg", tiles);

60	rectangle(img, Point(x,y), Point(x1,y1), Scalar(0,255,0), 1);

61

}

62

}

63

}

接下來，使用該imshow()函數(shù)顯示圖像補丁。imwrite()使用函數(shù) 將其保存到文件目錄中。

Python

1	#Save full image into file directory

2	cv2.imshow("Patched Image",img)

3	cv2.imwrite("patched.jpg",img)

4

5	cv2.waitKey()

6	cv2.destroyAllWindows()

C++

1	imshow("Patched Image", img);

2	imwrite("patched.jpg",img);

3	waitKey();

4	destroyAllWindows();

上面覆蓋有矩形補丁的最終圖像將如下所示： 

下圖顯示了保存到磁盤的單獨圖像補丁。