Unsharp Masking

Problem

Sharpen a grayscale image with unsharp masking.

Blur the image with a normalized 3x3 Gaussian kernel (sigma 1) using valid-mode cross-correlation, then set each center pixel to clamp(round(v + amount * (v - blur)), 0, 255), where v is the corresponding input pixel. The output shape is (H - 2) x (W - 2).

Hint

Add back a scaled version of the high-frequency detail (original minus blur).

Test cases

case 1

in: [[[4,4,4],[4,4,4],[4,4,4]],1]

expected: [[4]]

case 2

in: [[[0,0,0],[0,10,0],[0,0,0]],1]

expected: [[18]]

case 3

in: [[[1,2,3],[4,5,6],[7,8,9]],0.5]

expected: [[5]]

case 4

in: [[[1,2,3],[4,5,6],[7,8,9]],0]

expected: [[5]]

Starter codePython

def unsharp_mask(image, amount):
    # Your code here
    pass

4 test cases · run your code against them in the editor