Learning about opencv (1)

In the study opencv After basic operation of , We can expand it a little on the basis , Carry out some more advanced operations .

---

Catalog

One , Threshold image ：

Two , The image processing ：

1, Mean filtering ：

2, Box filtering ：

 3, Gauss filtering ：

4, median filtering ：

3、 ... and , morphology ：

1, Corrosion operation ：（ binary ： Like black and white ？）

2, Expansion operation ：（ Similar to corrosion ）

3, Open operation ：

4, Closed operation ：

5, Gradient operation ：

6, Hats and black hats ：

---

One , Threshold image ：

The first is about image threshold operation ：

cat = cv2.imread('cat.jpg')
#ret,dst = cv2.threshold(src,thresh,maxval,type)
# scr: Input diagram , Only single channel images can be input 
# dst: Output chart 
# thresh: threshold     Usually 127
# maxval: When the pixel exceeds the threshold （ according to type）, The value assigned 
# type: Binary operation , Five types ：
#cv2.THRESH_BINARY   Over threshold value maxval, Otherwise take 0
#cv2.THRESH_BINARTY_INV    cv.THRESH_BINARY The reversal of 
#cv2.THRESH_TRUNC     The part greater than the threshold value is set as the threshold value , Otherwise unchanged 
#cv2.THRESH_TOZERO    The part greater than the threshold does not change , Otherwise, it is set to 0
#cv2.THRESH_TOZERO_INV    cv.THRESH_TOZERO  The reversal of 

There are five basic types of threshold operations , The operation method is similar , Let's take the first example ：

ret,dst1 = cv2.threshold(cat,127,255,cv2.THRESH_BINARY)
cv_show('a',dst1)

                 

  The first is the original , The second is the image after the operation of the original image .

---

Two , The image processing ：

1, Mean filtering ：

Mean filtering is also called linear filtering , The main method is neighborhood average method . The basic principle of linear filtering is to replace each pixel value in the original image with the mean value , The current pixel to be processed （x,y）, Select a template , The template is composed of several neighboring pixels , Find the mean value of all pixels in the template , Then assign the mean value to the current pixel （x,y）, As the gray level of the processed image at this point g（x,y）, namely g（x,y）=∑f（x,y）/m m Is the total number of pixels in the template including the current pixel .

blur = cv2.blur(cat,(3,3))
cv_show('b',blur)

2, Box filtering ：

Similar to mean filtering , The difference is normalization .

# Box filtering ： Similar to the mean ( Normalization can be selected )
box1= cv2.boxFilter(cat,-1,(3,3),normalize=True)   # And divided by nine 
cv_show('c',box1)
# No need to normalize , Easy to cross the border 
box2= cv2.boxFilter(cat,-1,(3,3),normalize=False)
cv_show('d',box2)

            

  The first picture is normalized , Similar to the mean , The second graph is not normalized , It will be easy to cross the border .

 3, Gauss filtering ：

Gauss filter is a linear smoothing filter , Suitable for eliminating Gaussian noise , Noise reduction process widely used in image processing . Generally speaking , Gaussian filtering is the process of weighted average of the whole image . The specific operation of Gaussian filtering is ： Use a template （ Convolution 、 Mask ） Scan every pixel in the image , Using the weighted average gray value of the pixels in the neighborhood determined by the template to replace the value of the central pixel of the template .

# gaussian ： It's important to be close ,1 Said the weight ( Normal distribution )
aussian = cv2.GaussianBlur(cat,(5,5),1)
cv_show('e',aussian)

​ 

4, median filtering ：

Arrange the numbers in the matrix by size , Take the middle number as the value of the matrix .

# median filtering : Replace with a median 
median = cv2.medianBlur(cat,5)
cv_show('f',median)

  These methods are all for removing noise , Make the image clearer .

The following line of code is for the convenience of comparing the differences of various image operations .

res = np.hstack((blur,aussian,median))
cv_show('j',res)

​

---

3、 ... and , morphology ：

1, Corrosion operation ：（ binary ： Like black and white ？）

kernel = np.ones((5,5),np.uint8)         #（5,5） Range 
erosion = cv2.erode(cat,kernel,iterations=1)     #iterations: How many times 
cv_show('h',erosion)

​

These are execution 1,2,3 The result after the first time .

2, Expansion operation ：（ Similar to corrosion ）

kernel = np.ones((5,5),np.uint8)
dilate =cv2.dilate(cat,kernel,iterations=1)
cv_show('j',dilate)

​

  These are execution 1,2,3 The result after the first time .

Changing the scope can get different results , Take one execution as an example ： The smaller the range , The closer to the original , The larger the scope , The closer it is to the figure after multiple executions .

3, Open operation ：

Corrosion before expansion , Can deburr

# open ： Corrosion before expansion （ Deburring ？）
kernel = np.ones((5,5),np.uint8)
opening = cv2.morphologyEx(cat,cv2.MORPH_OPEN,kernel)
cv_show('i',opening)

4, Closed operation ：

Expand first and then corrode , It can mend holes

# close ： Expand first and then corrode （ Mend a hole ？）
kernel = np.ones((5,5),np.uint8)
closing = cv2.morphologyEx(cat,cv2.MORPH_CLOSE,kernel)
cv_show('i',closing)

​             ​

  On the left is the open operation , On the right is the closed operation .

5, Gradient operation ：

gradient = inflation - corrosion

kernel = np.ones((5,5),np.uint8)
dilate= cv2.dilate(cat,kernel,iterations=5)
cv_show('g',dilate)

​

6, Hats and black hats ：

formal hat = original - Open operation ：

# formal hat = original - Open operation 
tophat = cv2.morphologyEx(cat,cv2.MORPH_TOPHAT,kernel)
cv_show('k',tophat)

Black hat = Closed operation - original

tophat = cv2.morphologyEx(cat,cv2.MORPH_TOPHAT,kernel)
cv_show('k',tophat)

​​ 

  The first picture is a hat , The second picture is black hat .

---

That's all for this study , You are welcome to correct any mistakes .

（ notes ： For comparison , Lots of pictures are quoted ）