System Information

OpencCV python version: 4.7.072
Operating System: Windows 11 pro
Python version: 3.8.19

Detailed description

I am used to use cv2.norm(img1, img2) to compare images.
Using that function got slow with any opencv-python version newer than 4.2.0.34.
open-cv version 4.2.0.34 is the last one working fast.

I tested various opencv-python versions (4.10.0.84, 4.8.0.76, 4.7.072 with python 3.8.19 and also tested python 3.10.14 without any improvement.
But I need to stay with python 3.8.19 and opencv version >= 4.7 for compatibility reasons with other libs.

cv2.norm(img1 - img2) works faster than cv2.norm(img1, img2) on opencv-python v > 4.2.0.34 but is still slower than 4.2.0.34 with cv2.norm(img1, img2)

Even cv2.matchTemplate(img1, img2) is faster then cv2.norm(img, img2) on opencv-python v > 4.2.0.34

Any idea or fix to get the old speed back ?

Steps to reproduce

This is a small sample to show the speed difference.
My results:
using versions:
opencv-contrib-python 4.7.0.72
opencv-python 4.7.0.72
opencv-python-headless 4.7.0.72
"""
norm1c2 : cv2.norm( , ) : count = 1000
Real time: 1.45 seconds
CPU time: 0.08 seconds
Res: [2804.09022679371, 2804.09022679371, 0.0, 0.0]

norm1m2 : cv2.norm( - ) : count = 1000
Real time: 0.01 seconds
CPU time: 0.00 seconds
Res: [8761.552944541281, 7376.9984411005535, 0.0, 0.0]
"""

using versions
opencv-contrib-python 4.2.0.34
opencv-python 4.2.0.34
opencv-python-headless 4.2.0.34
"""
norm1c2 : cv2.norm( , ) : count = 1000
Real time: 0.00 seconds
CPU time: 0.00 seconds
Res: [2804.09022679371, 2804.09022679371, 0.0, 0.0]

norm1m2 : cv2.norm( - ) : count = 1000
Real time: 0.01 seconds
CPU time: 0.02 seconds
Res: [8761.552944541281, 7376.9984411005535, 0.0, 0.0]
"""

The sample program:

import cv2
import time

img1 = cv2.imread('./img1.png')
img2 = cv2.imread('./img2.png')
res = [0, 0, 0, 0]
cnt = 1000


def norm1c2():
    t1 = time.perf_counter(), time.process_time()
    for i in range(0, cnt):
        res[0] = cv2.norm(img1, img2)
        res[1] = cv2.norm(img2, img1)
        res[2] = cv2.norm(img1, img1)
        res[3] = cv2.norm(img2, img2)
    t2 = time.perf_counter(), time.process_time()

    print(f"\nnorm1c2 : cv2.norm( , ) : count = {cnt}")
    print(f" Real time: {t2[0] - t1[0]:.2f} seconds")
    print(f" CPU time: {t2[1] - t1[1]:.2f} seconds")
    print(f" Res: {res}")


def norm1m2():
    t1 = time.perf_counter(), time.process_time()

    for i in range(0, cnt):
        res[0] = cv2.norm(img1 - img2)
        res[1] = cv2.norm(img2 - img1)
        res[2] = cv2.norm(img1 - img1)
        res[3] = cv2.norm(img2 - img2)

    t2 = time.perf_counter(), time.process_time()
    print(f"\nnorm1m2 : cv2.norm( - ) : count = {cnt}")
    print(f" Real time: {t2[0] - t1[0]:.2f} seconds")
    print(f" CPU time: {t2[1] - t1[1]:.2f} seconds")
    print(f" Res: {res}")


norm1c2()
norm1m2()

img files attached

Issue submission checklist

• I report the issue, it's not a question
• I checked the problem with documentation, FAQ, open issues, forum.opencv.org, Stack Overflow, etc and have not found any solution
• I updated to the latest OpenCV version and the issue is still there
• There is reproducer code and related data files (videos, images, onnx, etc)