OpenCV 4 with Python Blueprints - Second Edition

Blending the images together

First, we create a MultiBandBlender object, which will help us merge images together. Instead of just picking values from one or the other image, blender will do interpolation between the available values:

    blender = cv2.detail_MultiBandBlender()    blend_width = np.sqrt(canvas_size[2] * canvas_size[3]) * 5 / 100    blender.setNumBands((np.log(blend_width) / np.log(2.) - 1.).astype(np.int))    blender.prepare(canvas_size)

Then, for each of the connected images, we do the following:

1. We warp the image and get the corner locations:

    for i, img in enumerate(conn_images):        K = cameras[i].K().astype(np.float32)        corner, image_wp = warper.warp(img, K, cameras[i].R,                                       cv2.INTER_LINEAR, cv2.BORDER_REFLECT)

2. Then, calculate the