[Python] Using OpenCV with Python (Basic)

Install (Anaconda)

To install the latest OpenCV 3.1 on Anaconda as of February 2016, get it from the following site.

conda install --channel https://conda.anaconda.org/menpo opencv3

Import and check the version.

In [1]: import cv2

In [2]: print cv2.__version__
3.1.0

Load image

Read the image with cv2.imread. At this time, the image is BGR. Please note that if you try to display with matplotlib as it is, the color will be strange.

When you read an image with sckit kimage or scipy, it becomes RGB.

The type of the loaded image is numpy array.

In [5]: I = cv2.imread('Lenna.bmp')

In [6]: I = cv2.imread('lena512color.tiff')

In [7]: type(I)
Out[7]: numpy.ndarray

For PIL

In [8]: from PIL import Image

In [9]: I = Image.open('Lenna.bmp')

In [10]: type(I)
Out[10]: PIL.BmpImagePlugin.BmpImageFile

For Scikit-image

In [11]: from skimage import io

In [12]: I = io.imread('Lenna.bmp')

In [13]: type(I)
Out[13]: numpy.ndarray

In [14]: Icv = cv2.imread('Lenna.bmp')

In [15]: print Icv[0,0,:]
[125 137 226]

In [16]: print I[0,0,:]
[226 137 125]

For Scypi

In [17]: from scipy import misc

In [18]: I = misc.imread('Lenna.bmp')

In [19]: type(I)
Out[19]: numpy.ndarray

In [20]: print I[0,0,:]
[226 137 125]

For matplotlib

In [21]: from matplotlib import image

In [22]: I = image.imread('Lenna.bmp')

In [23]: type(I)
Out[23]: numpy.ndarray

In [24]: print I[0,0,:]
[226 137 125]

Save the image

If you want to save the image with opencv, use imwrite.

imwrite(filename, img[, params])

In [26]: cv2.imwrite('output.png', I)
Out[26]: True

However, when saving with OpenCV, the colors must be arranged in the order of BGR in Interleaved format. If you save it as it is, it will be as follows.

Display image

Use imshow to display the image.

imshow(winname, mat)

import cv2

I = cv2.imread('./data/SIDBA/Lenna.bmp')

cv2.namedWindow('window')
cv2.imshow('window', I)
cv2.waitKey(0)
cv2.destroyAllWindows()

However, if you do this with interactive ipython, ipython will die.

When doing with ipython, it is good to execute startWindowThread first as shown below.

In [7]: I = cv2.imread('./data/SIDBA/Lenna.bmp')

In [8]: cv2.startWindowThread()
Out[8]: 1

In [9]: cv2.namedWindow('window')

In [10]: cv2.imshow('window', I)

In [11]: cv2.waitKey(0)
Out[11]: 10

In [12]: cv2.destroyAllWindows()

However, if you want to display an image, it is easier to use matplotlib unless you have a specific reason.

In [12]: import matplotlib.pyplot as plt
    ...: import cv2
    ...: 
    ...: I = cv2.imread('./data/SIDBA/Lenna.bmp')
    ...: 
    ...: plt.imshow(cv2.cvtColor(I, cv2.COLOR_BGR2RGB))
    ...: plt.show()

Color conversion

If you read the image with opencv, the image is BGR, so you need to convert it to RGB to display it with matplotlib. In such a case, use cvtColor.

import numpy as np
import cv2
import matplotlib.pyplot as plt

I = cv2.imread('./data/SIDBA/Lenna.bmp')

J = cv2.cvtColor(I, cv2.COLOR_BGR2RGB)

plt.imshow(J)
plt.show()