Foundation: 2 The essence of image

What is the essence of images ？–numpy And image Basics

Use jupyterLab Carry out preliminary study , First, install the software . As shown in the figure below ：

Installation steps

1. Create a python A virtual environment

conda create --name demo_py3.8 python=3.8

2. Install according to the source change .

3. function jupyter-lab, open , Then the URL will pop up .

http://localhost:8888/lab

Learn to use numpy

#  In order to understand the essence of image , We need to understand the concepts of arrays and matrices first 
#  First of all, remember , How to create a list 

list_a = [1,2,3,4,5]

#  Use type Command to view data types , yes list type 
type(list_a)
#  Create another list , Make the elements of this list still be lists 
list_b = [ [1,2], [3,4], [5,6]  ]

#  Print list  [[1, 2], [3, 4], [5, 6]]
list_b
#  By looking for a numerical index , Print page list_b The first 2 The number of elements 1 Elements  3
list_b[1][0]

#  This structure is also called array , such as list_a It's a one-dimensional array ,list_b Is a two-dimensional array 
#  For more efficient processing of arrays , We often use numpy In this bag 
#  First, import. numpy package , Rename it 
import numpy as np

#  that numpy How to create an array ？
#  It can be used Python List direct conversion 

#  First create a Python  list 
list_c = [1,2,3,4]

#  In the use of np.array() take Python The list is converted to numpy Array of  numpy.ndarray
my_array = np.array(list_c)
#  A print  array([1, 2, 3, 4])
my_array

#  that numpy There are also some built-in functions that can quickly create arrays 

#  For example, we use np.arange() You can quickly create an array of consecutive numbers 
#  For example, I create a 0~9  One dimensional array of  array([0, 1, 2, 3, 4, 5, 6, 7, 8, 9])
np.arange(0,10)

# jupyterlab Use in shift+tab You can view the help document of the function （ Check it out. , Input now np.arange(), Pop up corresponding function help document ）
#  You can see this arange() Function has start、stop and step Parameters , Each represents the starting value , Termination value , And step size 
#  If I want to create 0~10 An array of consecutive even numbers in , Just set the step size to 2（ Input now np.arange(0,10,2)）
# array([0, 2, 4, 6, 8])
np.arange(0,10,2)

#  You can also use np.ones It's all about 1 Array of ,（ When the input np.ones(), Pop up help notes ）
#  For example, I want to create a size of 3*3 Of all the 1 Array 

np.ones(shape=(3,3))
 #  Create another one with the size 10*3 Of all the 1 Array 
#  You can see 10 It's the number of lines ,3 It's the number of columns 

np.ones(shape=(10,3))
#  Or use np.zeros() whole 0 Array 
#  For example, the creation size is 5*5 Of all the 0 Array 

np.zeros(shape=(5,5))


#  use first np.randint Function some random integers  array([53, 86, 55, 17, 87, 12, 28, 95, 2, 44])
arr = np.random.randint(0,100,10)


#  Use max Get the maximum 
arr.max()
#  Reuse argmax()  Get the index of the maximum value 
arr.argmax()
#  Use min Function to get the minimum value 
arr.min()
#  Use argmin Get the minimum index 
arr.argmin()
#  Use mean() Method to get the average value 
arr.mean()
#  If you want to get humpy Size of array , Use numpy.shape, (10,)
arr.shape

#  You can also use reshape Function to convert the shape of an array , For example, I will arr convert to 5*2 Array of 
arr.reshape((5,2))
#  And then change into 2 That's ok 5 Column 
arr.reshape((2,5))
#  If you try to deform to a size of 2*10 Well ？ Will report a mistake ：ValueError: cannot reshape array of size 10 into shape (2,10)
arr.reshape((2,10))
#  That two-dimensional array , We also call it matrix in mathematics , Let's see numpy Operations on matrices 


#  First create a 10*10  Matrix 
matrix = np.arange(0,100).reshape((10,10))

#  Check the size ：(10, 10)
matrix.shape
#  Use the index method in brackets , Get the corresponding elements of the matrix , For example, I got the No 3 Xing di 5 Column elements 
matrix[2,4]
#  If you want to get all the elements in a row , We need to use numpy The section of ：
#  For example, I want to get the number 3 Row all elements , Just change the second position to a colon :
matrix[2,:]

#  Allied , For example, I want to get the number 6 Column all elements , Just change the first position to a colon :
matrix[:,5]
#  For example, I want to get the number 1~3 That's ok , The first 2~4 Column matrices , We can use numbers and colons to get 
matrix[0:3,1:4]

#  Of course, we can use the equal sign assignment statement , For example, I assign these positions 0
matrix[0:3,1:4] = 0
#  good , That's all numpy Operation usage of arrays and matrices 

A two-dimensional matrix is first and last , You can compare and understand why the image is calculated high first , In calculating the width , The essence is a three-dimensional matrix .

The cat figure understands the image

 #  First, import. numpy
import numpy as np
#  In order to be in notebook Show pictures in , Import matplotlib library 
import matplotlib.pyplot as plt
#  Add this line to Notebook Display images 
%matplotlib inline
#  Use one more PIL library , For reading images 
from PIL import Image
#  I am here img There is a picture under the folder （ Show me ）
#  We use it PIL Library reads images , Pay attention to the correct path 
img = Image.open('./img/cat.jpg')
#  Display images 
img

 #  Look at the types of variables ：PIL.JpegImagePlugin.JpegImageFile
type(img)

#  You can see that this is not numpy Array format of , that numpy Can't handle it yet 

#  First we need to translate it into numpy  Array , Use numpy.asarray() function 
img_arr = np.asarray(img)
#  View type ：numpy.ndarray
type(img_arr)

#  Let's look at the size ：(1253, 1880, 3)  Height , Width ,RGB Three channels 
img_arr.shape
#  Reuse matplot Of imshow() Method display Numpy An array of pictures 

plt.imshow(img_arr)
#  You can see that the abscissa and ordinate show that the length of the picture is 1800 many , Height is 1200 many 

#  Let's continue to operate on this picture , First use numpy Of copy Method to copy an original drawing 
img_arr_copy = img_arr.copy()
#  Check the size ：(1253, 1880, 3)
img_arr_copy.shape
#  use first numpy section , take R,G,B Of the three color channels R The red channel is displayed 
plt.imshow(img_arr_copy[:,:,0])

 #  You will find this color very strange , They are all emerald green , Why does it appear like this ？

#  We turn on matplot About the color table colormap Explanation ：
# https://matplotlib.org/stable/gallery/color/colormap_reference.html
#  You can see the default color ： It's Emerald （viridis ）. This color is convenient for color blind to watch 

#  We can also put cmap The color is set to volcanic magma style ：magma
plt.imshow(img_arr_copy[:,:,0],cmap='magma')

 #  Let's print the red R An array of channels 
img_arr_copy[:,:,0]
#  good , We know , The computer can't tell which channel is red , Each color channel is actually a grayscale image , Let's start with cmap The color is set to gray Grayscale 
#  to glance at 
plt.imshow(img_arr_copy[:,:,0],cmap='gray') 

#  We know , In the red channel  0 There is no red , It's pure black , And the closer it gets 255 Well , It means the more popular ,255 Just pure red 
#  Look at this grayscale image , The lighter the color , It means the more popular it is here  
#  We can take a look at the grayscale image of the red channel. The lightest color is this pendant （ Mouse indication ）
#  That goes back to the original color picture , You can see that this pendant is indeed the most popular 

#  Allied , We also show the green channel in grayscale mode 
plt.imshow(img_arr_copy[:,:,1],cmap='gray')