Computer vision: Pad an image three ways

Before filtering border pixels, a filter needs values outside the image. Build the three common choices.

What to build

A program that pads an image by p pixels on every side.

Requirements

  • zero fills the border with 0.
  • replicate repeats the nearest edge pixel (aaa, abc, ccc).
  • reflect mirrors the image without repeating the edge pixel (cb, abc, ba), as OpenCV does by default.
  • Print the padded image: H + 2p rows of W + 2p values.

Input

Line 1: H, W and p. Then H rows of W values. Last line: the mode.

Output

The padded rows.

Constraints

1 ≤ H, W ≤ 30; for reflect, p < H and p < W; otherwise p ≤ 10.

Example

Input:

2 2 1
1 2
3 4
zero

Output:

0 0 0 0
0 1 2 0
0 3 4 0
0 0 0 0

Why it works

Each padded position maps back to one real pixel (or to 0), so the padded image is built by working out that index for the row and the column separately.

Examples

Example 1 · Input
2 2 1
1 2
3 4
zero
Expected output
0 0 0 0
0 1 2 0
0 3 4 0
0 0 0 0
Example 2 · Input
2 3 1
1 2 3
4 5 6
replicate
Expected output
1 1 2 3 3
1 1 2 3 3
4 4 5 6 6
4 4 5 6 6

Prompts inside input() are ignored when checking, so input("n: ") is fine.

Hints

Hint 1

Loop r from −p to H + p − 1 and c from −p to W + p − 1, and work out which real pixel each position copies.

Hint 2

Replicate clamps an index into 0 to n − 1. Reflect maps i below 0 to −i and i of n or more to 2(n − 1) − i.

Solution

Show one correct solution (try it yourself first)
h, w, p = map(int, input().split())
img = [list(map(int, input().split())) for _ in range(h)]
mode = input().strip()


def where(i, n):
    if mode == "replicate":
        return min(max(i, 0), n - 1)
    if i < 0:
        return -i
    if i >= n:
        return 2 * (n - 1) - i
    return i


for r in range(-p, h + p):
    row = []
    for c in range(-p, w + p):
        if mode == "zero":
            row.append(img[r][c] if 0 <= r < h and 0 <= c < w else 0)
        else:
            row.append(img[where(r, h)][where(c, w)])
    print(" ".join(map(str, row)))

Lesson: Convolution and kernels →