Python OpenCV中的numpy与图像类型转换操作

Python OpenCV存储图像使用的是Numpy存储，所以可以将Numpy当做图像类型操作，操作之前还需进行类型转换，转换到int8类型

import cv2
import numpy as np
# 使用numpy方式创建一个二维数组
img = np.ones((100,100))
# 转换成int8类型
img = np.int8(img)
# 颜色空间转换，单通道转换成多通道， 可选可不选
img = cv2.cvtColor(img, cv2.COLOR_GRAY2BGR)
cv2.imwrite("demo.jpg", img)

补充知识：Python中读取图片并转化为numpy.ndarray()数据的6种方式

方式：                                        返回类型

OpenCV                                      np.ndarray
PIL                                               PIL.JpegImagePlugin.JpegImageFile
keras.preprocessing.image         PIL.JpegImagePlugin.JpegImageFile
Skimage.io                                  np.ndarray
matplotlib.pyplot                          np.ndarray
matplotlib.image                          np.ndarray

import numpy as np
import cv2
from keras.preprocessing.image import ImageDataGenerator, array_to_img, img_to_array, load_img
from PIL import Image
import skimage.io as io
import matplotlib.pyplot as plt
import matplotlib.image as mpig

'''
方式：   返回类型
OpenCV   np.ndarray
PIL    PIL.JpegImagePlugin.JpegImageFile
keras.preprocessing.image PIL.JpegImagePlugin.JpegImageFile
Skimage.io   np.ndarray
matplotlib.pyplot  np.ndarray
matplotlib.image  np.ndarray
'''

imagePath="E:/DataSet/test1/trainSet/bus/300.jpg"

'''
方式一：使用OpenCV
'''
img1=cv2.imread(imagePath)
print("img1:",img1.shape)
print("img1:",type(img1))
print("-"*10)

'''
方式二：使用PIL
'''
img2=Image.open(imagePath)
print("img2:",img2)
print("img2:",type(img2))
#转换成np.ndarray格式
img2=np.array(img2)
print("img2:",img2.shape)
print("img2:",type(img2))
print("-"*10)

'''
方式三：使用keras.preprocessing.image
'''
img3=load_img(imagePath)
print("img3:",img3)
print("img3:",type(img3))
#转换成np.ndarray格式,使用np.array(),或者使用keras里的img_to_array()
#使用np.array()
#img3=np.array(img2)
#使用keras里的img_to_array()
img3=img_to_array(img3)
print("img3:",img3.shape)
print("img3:",type(img3))
print("-"*10)

'''
方式四：使用Skimage.io
'''
img4=io.imread(imagePath)
print("img4:",img4.shape)
print("img4:",type(img4))
print("-"*10)

'''
方式五：使用matplotlib.pyplot
'''
img5=plt.imread(imagePath)
print("img5:",img5.shape)
print("img5:",type(img5))
print("-"*10)

'''
方式六：使用matplotlib.image
'''
img6=mpig.imread(imagePath)
print("img6:",img6.shape)
print("img6:",type(img6))
print("-"*10)

运行结果：

Using TensorFlow backend.
img1: (256, 384, 3)
img1: <class 'numpy.ndarray'>
----------
img2: <PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=384x256 at 0x249608A8C50>
img2: <class 'PIL.JpegImagePlugin.JpegImageFile'>
img2: (256, 384, 3)
img2: <class 'numpy.ndarray'>
----------
img3: <PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=384x256 at 0x2496B5A23C8>
img3: <class 'PIL.JpegImagePlugin.JpegImageFile'>
img3: (256, 384, 3)
img3: <class 'numpy.ndarray'>
----------
img4: (256, 384, 3)
img4: <class 'numpy.ndarray'>
----------
img5: (256, 384, 3)
img5: <class 'numpy.ndarray'>
----------
img6: (256, 384, 3)
img6: <class 'numpy.ndarray'>
----------

来源：https://blog.csdn.net/qq_31261509/article/details/94383575