Basic operation of Opencv

1. Image filling

First define the image display function:

def cv_show(name, img):
    cv2.imshow(name, img)
    cv2.waitKey(0)
    cv2.destroyAllWindows()

Image read:

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

Define the size of the image padding:

top_size, bottom_size, left_size, right_size = (50, 50, 50, 50)

 Next, use the copy method, reflection method, outer packaging method, and constant method to fill.

Copy method:

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

Reflection method: 

# 反射法，对感兴趣的图像中的像素在两边进行复制
reflect = cv2.copyMakeBorder(img_cat, top_size, bottom_size, left_size, right_size, borderType=cv2.BORDER_REFLECT)

second reflection method

# 反射法，也就是以最边缘像素为周，对称
reflect101 = cv2.copyMakeBorder(img_cat, top_size, bottom_size, left_size, right_size,
                                borderType=cv2.BORDER_REFLECT_101)

Packaging method:

# 外包装法
warp = cv2.copyMakeBorder(img_cat, top_size, bottom_size, left_size, right_size, borderType=cv2.BORDER_WRAP)

Constant method:

# 常量法，常数值填充，此处填充灰色
constant = cv2.copyMakeBorder(img_cat, top_size, bottom_size, left_size, right_size, borderType=cv2.BORDER_CONSTANT,
                              value=0)

The output results are compared as follows:

 

 2. Morphological operation

First read in the raw image:

pie = cv2.imread('pie.jfif')

 

 

Define the convolution kernel:

kernel = np.ones((10, 10), np.uint8)

Corrosion function:

dige_erosion = cv2.erode(pie, kernel=kernel, iterations=1)

The output is as follows:

 

Inflation function:

dige_dilate = cv2.dilate(pie, kernel=kernel, iterations=1)

The output is as follows:

Open operation: corrode first and then dilate

opening = cv2.morphologyEx(pie, cv2.MORPH_OPEN, kernel=kernel)

The output is as follows:

Closed operation: expand first and then corrode

closing = cv2.morphologyEx(pie, cv2.MORPH_CLOSE, kernel=kernel)

The output is as follows:

 

Gradient: Dilation-Erosion

gradient = cv2.morphologyEx(pie, cv2.MORPH_GRADIENT, kernel)

The output is as follows:

Top hat: original input - open result

tophat = cv2.morphologyEx(pie, cv2.MORPH_TOPHAT, kernel)

The output is as follows:

Black Hat: Closed Operation - Raw Input

blackhat = cv2.morphologyEx(pie, cv2.MORPH_BLACKHAT, kernel)

The output is as follows:

 3. Image smoothing

Pass in raw image:

Mean Filtering: A Simple Average Convolution Operation

blur = cv2.blur(cat, (3, 3))

 Box filtering: basically the same as the mean value, you can choose normalization, normalize when normalize is true, the same as the mean value filter

box = cv2.boxFilter(cat, -1, (3, 3), normalize=True)

 

 When normalize is False, normalization is not performed, and mean value processing is not performed, and 255 is taken after crossing the boundary

box = cv2.boxFilter(cat, -1, (3, 3), normalize=False)

 Gaussian filtering: The value in the convolution kernel of Gaussian blur satisfies the Gaussian distribution, which is equivalent to paying more attention to the middle

gaussian = cv2.GaussianBlur(cat, (3, 3), 1)

Median filtering: equivalent to replacing with median

median = cv2.medianBlur(cat, 3)