OpenCV最近邻居插值创建原始图像中不存在的像素

Question

I have an image A. The goal is to downsize it to an image, such that the dimensions of the image are smaller, but no pixel in the new image contains a value not present in the original. In other words, if $p \\in A'$ then $p \\in A$.

I tried doing this in python as follows:

resized = cv2.resize(old_img, (1024, 1024), 0, 0, cv2.INTER_NEAREST)

However, when the input is a binary image (old_img contains only black and white pixels), the resulting image contains grey values (values not equal to 0 or 255).

Is this a bug in the implementation? It's a hard requirement for me that no pixel ends up with a non original value.

Answer 1

I believe it should be

resized = cv2.resize(old_img, (1024, 1024), 0, 0, interpolation = cv2.INTER_NEAREST)

Note the interpolation = ... as part of the 3rd argument.

See here