OpenCV is a powerful tool to process images. In this tutorial, we will introduce how to read an image to numpy ndarray.
Python pillow library also can read an image to numpy ndarray.
Python Pillow Read Image to NumPy Array: A Step Guide
Preliminary
We will prepare an image which contains alpha chanel. We will start to read it using python opencv.
This image is (width, height)=(180, 220), the backgroud of it is transparent.
Read image using python opencv
Import library
import cv2 import numpy as np
Read image
We can use cv2.imread() to read an image. This function is defined as:
cv2.imread(path ,flag)
path: the path of image
flag: determines the mode of reading an image
These are some values are often used in flag:
| flag | Description |
| cv2.IMREAD_COLOR | Load the image in BGR color format, ignore alpha channel |
| cv2.IMREAD_UNCHANGED | Similar to cv2.IMREAD_COLOR. Including alpha channel if exist |
| cv2.IMREAD_GRAYSCALE | Loads image in grayscale mode |
We will use an example to show you how to do.
flag = [cv2.IMREAD_COLOR, cv2.IMREAD_UNCHANGED, cv2.IMREAD_GRAYSCALE]
for f in flag:
img = cv2.imread('opencv.png', f)
print(type(img))
print(img.shape)
print(img)
cv2.imshow('image window',img)
cv2.waitKey(0)
cv2.destroyAllWindows()
Run this code, we can find:
img is <class ‘numpy.ndarray’>
| flag | shape | Data | Image |
| cv2.IMREAD_COLOR | (222, 180, 3) | [[[0 0 0] [0 0 0] [0 0 0] … [0 0 0] [0 0 0] [0 0 0]]] |
 |
| cv2.IMREAD_UNCHANGED | (222, 180, 4) | [[[0 0 0 0] [0 0 0 0] [0 0 0 0] … [0 0 0 0] [0 0 0 0] [0 0 0 0]]] |
 |
| cv2.IMREAD_GRAYSCALE | (222, 180) | [[0 0 0 … 0 0 0] [0 0 0 … 0 0 0] [0 0 0 … 0 0 0] … [0 0 0 … 0 0 0] [0 0 0 … 0 0 0] [0 0 0 … 0 0 0]]] |
 |
Notice:
If you read an image which does not contains alpha channel (it is not transparent) using cv2.IMREAD_UNCHANGED, you may get an shape(heigh, width, 3)
For example:
As to this image, it is not transparent.
We will read it using cv2.IMREAD_UNCHANGED. You will get:
<class 'numpy.ndarray'> (367, 384, 3) [[[255 255 255] [255 255 255] [255 255 255] ... [255 255 255] [255 255 255] [255 255 255]]]
The effect of cv2.IMREAD_UNCHANGED is same to cv2.IMREAD_COLOR.