CODEWITHSUNDEEP
pythonopencvrgba
Tom
Tom

Reputation: 15

RGBA transparency not outputting as clear when overlaid onto image

I have got a script to spot a pothole in an image and draw an ellipse around it. I have two images. one the original image and then a separate numpy array which is white with a red ellipse on it. I have converted the ellipse image RGBA image with the intention of overlaying it onto the original image, using the code below. However the output image ( new_img ) is showing a blend with the ellipse image background still showing at half opacity (because the transparency is set in blend() however I would like the white background to be completely clear and the ellipse to be half transparent) . In other words the for loop is not creating a RGBA image with alpha of 0. Attached are the input, ellipse image and output.

from PIL import Image
import cv2

# Make True pixels red
RGB[test_ellipse==255]  = [255,0,0]
# Make False pixels blue
RGB[test_ellipse==1] = [255,255,255]

image = cv2.imread(file_path , cv2.IMREAD_GRAYSCALE)
test = Image.fromarray(RGB)
over = test.convert('RGBA')
datas = over.getdata()


newData = []
for item in datas:
    if item[0] == 255 and item[1] == 255 and item[2] == 255:
        newData.append((255, 255, 255, 0))
    else:
       newData.append(item)

over.putdata(newData)
background = Image.fromarray(image)
background = background.convert('RGBA')
new_img = Image.blend(background, over, 0.5)
new_img.show()

Input ellipse overlay Current output

Upvotes: 0

Views: 763

Answers (1)

Mark Setchell
Mark Setchell

Reputation: 207465

Hopefully this will give you an idea of adding an alpha channel. I just mocked up a dummy alpha channel with a painting program, making it mid-grey with one black and one white circle:

enter image description here

#!/usr/bin/env python3

import cv2
import numpy as np

# Load image and alpha
im    = cv2.imread('pothole.png', cv2.IMREAD_GRAYSCALE)
alpha = cv2.imread('alpha.png',   cv2.IMREAD_GRAYSCALE)

# Make grey pothole picture 3 channel grey and add alpha layer
res = np.dstack((im,im,im,alpha))
cv2.imwrite('result.png', res)

enter image description here

You can maybe see the effect more clearly if I undertile with a chessboard like Photoshop does:

enter image description here

Upvotes: 1

Related Questions