Possible bug / precision issue in normalize() about opencv HOT 2 OPEN

I think there is no bug in normalize

import numpy as np
import cv2 as cv


data = np.array([0, 2, 20, 103., 240., 15.], dtype=np.float32)
norm_opencv = cv.normalize(data, None, 0, 255, cv.NORM_MINMAX)
dMin=0
dMax =255
smax = np.max(data)
smin = np.min(data)
scale = (dMax- dMin)* 1 /(smax - smin )
shift = dMin - smin*scale
norm_numpy = data * scale + shift
print(np.max(np.abs(norm_opencv-norm_numpy)))

Result 0.0

if something is weird it is in python binding for np.array to Mat or UMat

from opencv.

Hello, thank you for the reply. Indeed with the values from your example there is no difference. However when I change your example values to the values from my example, there is the same difference that I reported in the beginning. Could you please double check that this happens with my values on your side as well?

from opencv.