我如何在 0 到 26 之间更快地标准化像素

Question

I have a vector with about 150528 pixels.

img1 = image.load_img(path = "image.jpg", target_size = (224, 224, 3))

img1_pixel_array = image.img_to_array(img = img1, dtype = np.uint8)

a = img1_pixel_array.reshape(150528,)

I normalizing between 0 and 26 for this data.

list = []
for i in a:
    z = (i - min(a)) / (max(a) - min(a)) * 26
    list.append(z)

The problem is that only 1000 pixels are processed in 40 seconds. My question is how do I normalize this data between 0 and 26 faster.

Answer 1

If seems that you are recomputing the minimum (twice) and maximum of a on every iteration. This does not sound like a good idea.

You'd better precompute the scaling coefficients A and B such that z = A i + B.