CODEWITHSUNDEEP
pythonopencvimage-processing
mefahimrahman
mefahimrahman

Reputation: 349

How to determine the number of channels in image?

I want to see the number of channels for thermal images, RGB images, grayscale images and binary images.

So I write this program:

import cv2
import numpy

img = cv2.imread("B2DBy.jpg")
print('No of Channel is: ' + str(img.ndim))

cv2.imshow("Channel", img)
cv2.waitKey()

But it gives the same three channel results for all types of images? I've read this question but it gives an error:

img = cv2.imread("B2DBy.jpg", CV_LOAD_IMAGE_UNCHANGED)
NameError: name 'CV_LOAD_IMAGE_UNCHANGED' is not defined

So my question is: Is it is the right way to see the number of channels? Or, somehow, I entered three channel images all the time and thus it gives three channel output?

My inputs:

Thermal Image

Gray sacle Image

Binary Image

RGB Image

Upvotes: 1

Views: 14553

Answers (3)

HansHirse
HansHirse

Reputation: 18925

The correct parameter in your cv2.imread should be:

img = cv2.imread('path/to/your/image', cv2.IMREAD_UNCHANGED)

Let's have a look at your images now. I use ImageJ's Show Info... command as well as the following Python code with OpenCV and Pillow:

import cv2
from PIL import Image

img_pil = Image.open('path/to/your/image')
print('Pillow: ', img_pil.mode, img_pil.size)

img = cv2.imread('path/to/your/image', cv2.IMREAD_UNCHANGED)
print('OpenCV: ', img.shape)

First image (depth map)

Pillow:  RGB (640, 512)
OpenCV:  (512, 640, 3)

ImageJ also says, that's a RGB image. So, most likely, your depth map was just saved as a RGB png.

Second image (dog)

Pillow:  RGB (332, 300)
OpenCV:  (300, 332, 3)

Interestingly, ImageJ says, that's an grayscale jpg! I assume, OpenCV and Pillow just don't support grayscale jpg, although there seems to be a grayscale jpg format.

Third image (sign)

Pillow:  1 (200, 140)
OpenCV:  (140, 200)

Both, Pillow and OpenCV say, that's a grayscale image, which is also supported by ImageJ. Furthermore, Pillow uses mode '1' here, which is reflected by the dithered look of the image.

Fourth image (colours)

Pillow:  RGB (500, 333)
OpenCV:  (333, 500, 3)

That's just some RGB image; ImageJ also says this.

Conclusion

Yes, most likely, most of your images may just be RGB images. Nevertheless, using cv2.IMREAD_UNCHANGED at least will properly identify grayscale png files. It's questionable, if grayscale jpg files are properly supported.

Hope that helps!

----------------------------------------
System information
----------------------------------------
Platform:    Windows-10-10.0.16299-SP0
Python:      3.8.1
OpenCV:      4.2.0
Pillow:      7.0.0
----------------------------------------

Upvotes: 6

randy sandy
randy sandy

Reputation: 99

If image is grayscale you will need to set a flag, tuple returned contains only number of rows and columns.

So it is a good method to check if loaded image is grayscale or color image.

image = cv2.imread('gray.jpg', cv2.IMREAD_GRAYSCALE)
image.shape

If len(img.shape) gives you three, third element gives you number of channels.

Upvotes: 1

steak_Overcooked
steak_Overcooked

Reputation: 111

I'm not sure if it'll work, but documentation says this: cv.LoadImage(filename, iscolor=flag) with flags given. There's a flag "<0" that stands for "Return the loaded image as is (with alpha channel)". I would try this:

img = cv2.imread("B2DBy.jpg",iscolor=<0)

or this

img = cv2.imread("B2DBy.jpg",iscolor=CV_LOAD_IMAGE_ANYDEPTH)

Upvotes: 0

Related Questions