Peak Signal to Noise Ratio (PSNR) in Python for an Image

Question

How do I calculate the maximum signal to noise ratio (PSNR) in Python? Is there any library that can calculate PSNR for Image?

You can use python ski-image library. Using it you can calculate PSNR Value. — RUC..., Jan 17 '19 at 09:44
This does not provide an answer to the question. Once you have sufficient reputation you will be able to comment on any post; instead, provide answers that don't require clarification from the asker. - From Review — A_A, Jan 17 '19 at 10:33

score 11 · Answer 1 · answered Oct 25 '19 at 10:37

11

You can use cv2.PSNR like this example:


import cv2
img1 = cv2.imread('img1.bmp')
img2 = cv2.imread('img2.bmp')
psnr = cv2.PSNR(img1, img2)

answered Oct 25 '19 at 10:37

Amir Khakpour

111
1
2

score 8 · Answer 2 · edited Jun 22 '23 at 16:47

8

The steps for calculation of PSNR value of two images:

import math
import cv2
import numpy as np
original = cv2.imread("original.png")
contrast = cv2.imread("photoshopped.png", 1)
def psnr(img1, img2):
    mse = np.mean(np.square(np.subtract(img1.astype(np.int16),
                                        img2.astype(np.int16)))
    if mse == 0:
        return np.Inf
    PIXEL_MAX = 255.0
    return 20 * math.log10(PIXEL_MAX) - 10 * math.log10(mse)
d = psnr(original, contrast)
print(d)

Wiki

edited Jun 22 '23 at 16:47

lennon310

3,590
19
24
27

answered Jul 21 '18 at 03:23

Himanshu Tyagi

81
1
2

1

Why return 100 if the MSE == 0? – rj487 Apr 01 '19 at 06:20

score 5 · Answer 3 · answered Oct 21 '19 at 21:16

5

turn to float first!!!!!!!! turn to float first!!!!!!!! turn to float first!!!!!!!!

def compute_psnr(img1, img2):
img1 = img1.astype(np.float64) / 255.
img2 = img2.astype(np.float64) / 255.
mse = np.mean((img1 - img2) ** 2)
if mse == 0:
    return "Same Image"
return 10 * math.log10(1. / mse)

answered Oct 21 '19 at 21:16

Shuai Yan

51
1
1

Technically, you don't need to convert to float specifically. That is done automatically: np.mean(np.arange(3)**2) == 1.666.... However, if (img1 - img2) is of dtype np.uint8, you will of course not be able to store the results of the computation in an 8-bit unsigned integer. – Mateen Ulhaq Aug 11 '21 at 08:34

score 2 · Answer 4 · edited Apr 13 '17 at 12:47

2

If you have a replica of your signal (image) that is noise free, you can calculate the correlation coefficient which is directly related to SNR.

See my response here for specific details on determining the correlation coefficient and from that SNR:

Noise detection

In this context there is no "maximum SNR" but will be the SNR for your entire image, meaning the power of your desired signal relative to everything else (distortions).

edited Apr 13 '17 at 12:47

Community

1

answered Mar 02 '17 at 19:21

Dan Boschen

50,942
2
57
135

thanks for your reply. I posted that is there have any lib in python that can calculate PSNR?? – Sudip Das Mar 02 '17 at 19:26
1

I do not know of one that makes it any simpler than that- what I described is what I do in Python, and the approach is very straightforward. – Dan Boschen Mar 02 '17 at 19:29

Peak Signal to Noise Ratio (PSNR) in Python for an Image

4 Answers4

Linked