opencv threshold problem for finding diff images

Question

I have 2 images like below. I want to have differences two of them.

I tried some codes with threshold. But no threshold different. two images threshold image is all black. How can I make differences are with color white on thresh?

(the difference is just on top-left like a small dark)

before = cv2.imread('1.png')
after = cv2.imread('2.png')

threshb = cv2.threshold(before, 0, 255, cv2.THRESH_BINARY_INV)[1]
thresha = cv2.threshold(after, 0, 255, cv2.THRESH_BINARY_INV)[1]

cv.imshow("before",threshb)
cv.imshow("after",thresha)

Not: I used "structural_similarity" link here for finding differences but it founds a lot of differences :(

I don't need small pixel differences. I need differences like seen with human eyes.

Use cv2.absdiff(). See https://docs.opencv.org/4.1.1/d2/de8/group__core__array.html#ga6fef31bc8c4071cbc114a758a2b79c14 — fmw42, Aug 08 '21 at 00:52
show what is "didn not work" in your estimation. I'd suggest opencv's background subtraction module but that requires video. absdiff() should certainly solve this. — Christoph Rackwitz, Aug 08 '21 at 10:51
When I use absdiff with first and second pics, the result is all black — nec, Aug 08 '21 at 12:43

fmw42 · Accepted Answer · 2021-08-08T16:05:37.720

The way to handle that is to do absdiff followed by some gain in Python/OpenCV. When you do the absdiff with these images, the difference is small so will not be much above black. So you have to increase the gain to see the difference. (Or you could set an appropriate threshold)

Input 1:

Input 2:

import cv2
import numpy as np
from skimage import exposure as exposure

# read image 1
img1 = cv2.imread('gray1.png', cv2.IMREAD_GRAYSCALE)

# read image 2
img2 = cv2.imread('gray2.png', cv2.IMREAD_GRAYSCALE)

# do absdiff
diff = cv2.absdiff(img1,img2)

# apply gain
result1 = cv2.multiply(diff, 5)

# or do threshold
result2 = cv2.threshold(diff, 10, 255, cv2.THRESH_BINARY)[1]

# save result
cv2.imwrite('gray1_gray2_diff1.png', result1)
cv2.imwrite('gray1_gray2_diff2.png', result2)

# display result
cv2.imshow('diff', diff)
cv2.imshow('result1', result1)
cv2.imshow('result2', result2)
cv2.waitKey(0)
cv2.destroyAllWindows()

Result 1 (gain):

Result 2 (threshold):

opencv threshold problem for finding diff images

1 Answers1