在 Python 中重新映射 OpenCV 中的像素值的最快方法是什么？

Question

I have a single channel image gray_image , with pixels values [0.. 255] and a lookup table mapping those values to colors:

lookup = {
 0: [0, 0, 0],
 1: [23, 54, 35],
 ...
 255: [200, 52, 20],
}

What is the fastest way to create a new 3 channel image where each pixel is colored based on the lookup of its value in the original images, eg

color_image[y, x] = lookup[gray_image[y, x]]

Instead of iterating through every pixel and setting it individually?

Answer 1

Assuming that the keys of the lookup table are properly ordered, you could convert the dictionary into an array and then apply NumPy's advanced indexing like this:

import numpy as np

palette = np.array([row for row in lookup.values()])
color_image = palette[gray_image]

If the keys of your dictionary are not ordered, the code above won't work. In that case you could convert the dictionary into an array as follows:

palette = np.zeros(shape=(256, 3), dtype=np.uint8)
for key in lookup.keys():
    palette[key] = lookup[key]

This approach implicitly defines a default value of [0, 0, 0] , ie if the lookup table does not contain a certain index, let's say n , then those pixels with gray level n will be mapped to [0, 0, 0] (black).

Answer 2

palette = np.array([lookup[index] for index in range(256)], dtype=np.uint8)
color_image = palette[gray_image]

the other answer didn't make sure to sort the dictionary by key (python dicts only recently got a stable order of pairs) and it didn't make sure that all index values exist.