Python: Blend two different sized images using OpenCV

Question

I am trying to blend two images together after warping them. The problem, is that the images are not the same sizes. I have tried this to make the blend work:

#warped_image and image_2 are parameters, warped_image is always larger
output_image = np.copy(warped_image)

topBottom = (warped_image.shape[0] - image_2.shape[0]) // 2
leftRight = (warped_image.shape[1] - image_2.shape[1]) // 2

if warped_image.shape[0] % 2 != 0:
    warped_image = warped_image[0: warped_image.shape[0] - 1, 0: warped_image.shape[1] - 1]


image_2 = cv2.copyMakeBorder(image_2, topBottom, topBottom, leftRight, leftRight, cv2.BORDER_CONSTANT, value=[0,0,0])

while image_2.shape[0] != warped_image.shape[0]:
    image_2 = cv2.copyMakeBorder(image_2, 1, 1, 1, 1, cv2.BORDER_CONSTANT,
                                 value=[0, 0, 0])

cv2.addWeighted(warped_image, .5, image_2, .5, 0.0, output_image)

The problem is that the blend is inaccurate on the first attempt, and if I try to blend a already blended image with another one, the image becomes completely distorted.

Is there a better way to blend two images together at a specific point?

Edit Results:

Two images:

Blending a third to the first two:

The second image is actually too large to load in here. Its end size is 62521 × 4111, which is incredibly large, since the images are all 3200 × 2368

It should be able to blend. I was able to make a basic blend, where I just combined the images at a specific point and that worked — Soatl, Jun 22 '16 at 17:35
have a look at my recent answer in http://stackoverflow.com/questions/37911062/how-to-obtain-the-right-alpha-value-to-perfectly-blend-two-images/37918596#37918596 — Micka, Jun 22 '16 at 17:37
joeforker was making a joke referencing the Blendtec videos which ask the question "Will it blend?" — SSteve, Jun 23 '16 at 23:54

Python: Blend two different sized images using OpenCV

0 Answers0