Is there a faster way to judge whether two pictures are different?

Question

I am using it to tell whether now screenshot is different from last screenshot. Now I use

with open('last_screenshot.bmp','rb+') as f:
    org = f.read()
with open('now_screenshot.bmp','rb+') as f:
    new = f.read()
if(org==new):
    print("The pictures are same")

Is there a faster way to do this?

You want to know if they are exactly the same? Does this answer your question?: [link](https://stackoverflow.com/q/23195522/11261546) — Ivan, Jan 17 '20 at 13:41
@nathancy This is simalar to ```PIL.ImageChops.difference```.But what I want to do is to know whether they are different instead of where is different between them.Thanks for response. — jizhihaoSAMA, Jan 18 '20 at 04:18
@jizhihaoSAMA yes SSIM returns a difference mask, so you can check the mask to determine if they are different. If there are white pixels on the resulting mask, then the two images are different. SSIM is a possible method which can help you to determine IF they are different and where they are different — nathancy, Jan 20 '20 at 20:52
There is a chance that [template matching](https://docs.opencv.org/master/d4/dc6/tutorial_py_template_matching.html) might be marginally faster, but ` == ` seems like it should be good. — bfris, Jul 02 '20 at 23:59

score 1 · Answer 1 · answered Jan 17 '20 at 14:18

1

You won't get nowhere comparing pixels. Your options:

extract features using descriptor (HOG, SIFT, SURF, ORB), match them, see how many were matched; example
calculate hashes, compute hamming metric here's example
take pretrained embedder; when it comes to images it's pretty easy, just take activation of penultimate layer; it can be inception, vgg etc

answered Jan 17 '20 at 14:18

Piotr Rarus

884
8
16

Oh....bro,thanks.But I think this may be a little complex.However,your ideas make me learn a lot. – jizhihaoSAMA Jan 17 '20 at 14:34

score 0 · Answer 2 · answered Jan 17 '20 at 13:44

0

Well you could iterate the files chunk by chunk instead of reading the entire thing in memory.

Alternatively, use filecmp or shell out to cmp(1).

answered Jan 17 '20 at 13:44

Masklinn

34,759
3
38
57

score 0 · Answer 3 · answered Jan 17 '20 at 13:57

You can use filecmp.cmp(..., shallow=False) which ships with the standard library. This won't read the whole files into memory but instead read them in chunks and short-circuit when a differing chunk is encountered. Example usage:

import filecmp

if filecmp('last_screenshot.bmp', 'now_screenshot.bmp', shallow=False):
    print('Files compare equal')

Is there a faster way to judge whether two pictures are different?

3 Answers3