有没有什么简单的方法可以用 cv2 分析透明的 png 图像？

Question

I am trying to load a transparent image (just without background) with CV2 but when I try to show the image I get a black background. Is there any possible (and easy) way to load transparent PNG image? The input:

The result:

This is the result I get: Even if i'm trying this code:

import cv2
im = cv2.imread("image.png", cv2.IMREAD_UNCHANGED)
cv2.imshow('test',im)

I get the result with black background

Answer 1

The image you are loading has four channels. Check the array size. Channel 4 is the alpha channel and you can save the image back accordingly with an alpha channel. You can also generate a mask from the alpha channel (in OpenCV masks are binary with values 0, and 255, respectively).

Most of OpenCV operations work on 4-channel images, but they do not treat the alpha channel special or derive any information from it.

Answer 2

cv.loadImage() is for MultiCameraCalibration. Use cv.imread() for single images. To get transparency, you must use the flag of cv.IMREAD_UNCHANGED

Here is how to do that in Python/OpenCV. Note that with the cv.imshow you will not see transparency. It will be black. To do further analysis, you should extract the alpha channel as a mask and use that for whatever you want to do to limit the processing to the shoe and not the background around it.

import cv2

# load image with alpha channel
img = cv2.imread('shoe.png', cv2.IMREAD_UNCHANGED)

# get mask from alpha channel
mask = img[:,:,3]

# view images
# NOTE: imshow does not show transparency. It will be black there.
cv2.imshow('img', img)
cv2.imshow('mask', mask)
cv2.waitKey(0)
cv2.destroyAllWindows()

# Saving the image does contain transparency as can be seen when viewed with some external viewer.
cv2.imwrite('shoe_copy.png', img)