Otsu Threshold

Problem

Compute the Otsu threshold of a grayscale image.

Build a 256-bin histogram and choose the threshold t that maximizes the between-class variance w_b * w_f * (mean_b - mean_f)^2, where the classes are pixels <= t and pixels > t. Return the smallest such t; if the image is constant, return 0.

Hint

Sweep t from 0 to 255 while tracking the running sum of the low class.

Test cases

case 1

in: [[[0,0],[255,255]]]

expected: 0

case 2

in: [[[1,2],[3,4]]]

expected: 2

case 3

in: [[[0,1],[8,9]]]

expected: 1

case 4

in: [[[7,7],[7,7]]]

expected: 0

Starter codePython

def otsu_threshold(image):
    # Your code here
    pass

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