如何使用透明背景的opencv将png图像固定到另一个图像上

Question

I want to add glasses PNG image onto another image, but i get the PNG image with white background which hides the face.

I need to know what i did wrong please.

Original face image

Original glasses image

Expected result

import cv2
import numpy
img = cv2.imread("barack-obama.jpg")
lunette = cv2.imread("lunette.png", cv2.IMREAD_UNCHANGED)
eye_cascade = cv2.CascadeClassifier('haarcascade_eye.xml')
face_cascade = cv2.CascadeClassifier('haarcascade_frontalface_default.xml')

eyes = eye_cascade.detectMultiScale(img, scaleFactor = 1.1, minNeighbors = 5)   

r = 500.0 / img.shape[1]
dim = (500, int(img.shape[0] * r))
resized = cv2.resize(img, dim, interpolation = cv2.INTER_AREA)

 grey = cv2.cvtColor(resized, cv2.COLOR_BGR2GRAY)

faces = face_cascade.detectMultiScale(grey, 1.3, 5)

for (x,y,w,h) in faces:

roi_grey = grey[y:y+h, x:x+w]
roi_color = resized[y:y + h, x:x + w]
eyes = eye_cascade.detectMultiScale(roi_grey)
for (ex,ey,ew,eh) in eyes:
cv2.rectangle(roi_color,(ex,ey),(ex+ew,ey+eh),(0,255,0),2)

lunette = cv2.resize(lunette , (w,h))
w, h, c = lunette.shape 
for i in range(0, w):
for j in range(0, h):

    if lunette[i, j][0] != 0:
        resized[y + i, x + j] = lunette[i, j][1]
#help please
cv2.imshow('img',resized)

cv2.waitKey(0)

Answer 1

You can use cvzone library for solve this problem.

1. install cvzone library

   pip install cvzone

2. import cvzone in your project

    import cvzone

3. imread your emoji with following format

   lunette = cv2.imread("lunette.png", cv2.IMREAD_UNCHANGED)

4. now replace the your emoji like this:

    YOUR BACKGROUND PICTURE = cvzone.overlayPNG(YOUR BACKGROUND PICTURE, lunette , [x, y])

Answer 2

Here is one way to do that in Python/OpenCV/Numpy:

• Read the background image (Obama)
• Read the overlay image (glasses) unchanged
• Extract the base BGR channels from the overlay image
• Extract the alpha channel from the overlay image as a mask
• Insert the BGR channels of the overlay into the background image at the desired location
• Insert the mask into a black image the size of the background image at the desired location
• Use Numpy where to composite the new background and overlay images using the new mask image
• Save the result

import cv2
import numpy as np

# read background image
img = cv2.imread("obama.jpg")
ht, wd = img.shape[:2]

# read overlay image
img2 = cv2.imread("sunglasses.png", cv2.IMREAD_UNCHANGED)
ht2, wd2 = img2.shape[:2]

# extract alpha channel as mask and base bgr images
bgr = img2[:,:,0:3]
mask = img2[:,:,3]

# insert bgr into img at desired location and insert mask into black image
x = 580
y = 390

bgr_new = img.copy()
bgr_new[y:y+ht2, x:x+wd2] = bgr

mask_new = np.zeros((ht,wd), dtype=np.uint8)
mask_new[y:y+ht2, x:x+wd2] = mask
mask_new = cv2.cvtColor(mask_new, cv2.COLOR_GRAY2BGR)

# overlay the base bgr image onto img using mask
result = np.where(mask_new==255, bgr_new, img)

# save results
cv2.imwrite('obama_glasses.jpg', result)

# display results
cv2.imshow('bgr', bgr)
cv2.imshow('mask', mask)
cv2.imshow('bgr_new', bgr_new)
cv2.imshow('mask_new', mask_new)
cv2.imshow('result', result)
cv2.waitKey(0)
cv2.destroyAllWindows()

Result: