Skip to content
Search

Image Comparison

lastmod 2025-01-03created 2025-01-031 edits

Notes on image comparison.

Comparing Two Images in GUI Applications

Recommended methods for improving readability and efficiency when comparing two images and displaying them in a GUI.

There are various approaches depending on the purpose of the comparison (difference analysis, similarity checking, etc.), but the main methods are as follows.

Side-by-Side Comparison (left-right or top-bottom comparison)

  • Method: Place the two images next to each other. Arrange them left-right or top-bottom so the user compares the differences visually.
  • Advantage: The most intuitive, and it is easy to see the changed parts directly.
  • Implementation: Place the two images on a canvas with OpenCV, PyQt, Tkinter, etc.
  • Additional features:
    • Use a slider to adjust the transparency of one image, or zoom in on a specific region.
    • Use the mouse to sync positions between the images and compare the same point at the same time. 

import cv2
import numpy as np
from tkinter import *
from PIL import Image, ImageTk

# 이미지 불러오기
img1 = cv2.imread('image1.jpg')
img2 = cv2.imread('image2.jpg')

# 크기 조정 (동일 크기로)
img1 = cv2.resize(img1, (600, 400))
img2 = cv2.resize(img2, (600, 400))

# GUI 설정
root = Tk()
root.title("Image Comparison")

# 이미지 변환
img1_tk = ImageTk.PhotoImage(image=Image.fromarray(cv2.cvtColor(img1, cv2.COLOR_BGR2RGB)))
img2_tk = ImageTk.PhotoImage(image=Image.fromarray(cv2.cvtColor(img2, cv2.COLOR_BGR2RGB)))

# 라벨 배치
Label(root, image=img1_tk).grid(row=0, column=0)
Label(root, image=img2_tk).grid(row=0, column=1)

root.mainloop()

Difference Overlay

  • Method: Compare the two images pixel by pixel and highlight the differences. Mark the differing parts in a color (red, etc.) to clearly reveal the differences between the two images.
  • Advantage: Differences can be checked quickly, and subtle changes are detected effectively.
  • Implementation: Compute the difference between the images with cv2.absdiff() or cv2.subtract().
  • Additional features:
    • Treat only values above a certain threshold as differences and remove noise.
    • Add labels or borders to the differing parts. 

diff = cv2.absdiff(img1, img2)
gray_diff = cv2.cvtColor(diff, cv2.COLOR_BGR2GRAY)
_, thresh = cv2.threshold(gray_diff, 30, 255, cv2.THRESH_BINARY)

# 차이 부분 강조
highlight = cv2.addWeighted(img1, 0.6, img2, 0.6, 0)
highlight[thresh > 0] = (0, 0, 255)

cv2.imshow("Difference Highlight", highlight)
cv2.waitKey(0)
cv2.destroyAllWindows()

Slider Blending

  • Method: Compare by moving a slider to adjust the transparency between the two images. Dragging the slider to one side gradually transitions one image into the other.
  • Advantage: Differences can be checked in real time, and the user can directly inspect the parts they want.
  • Implementation: Implement a slider with OpenCV and Tkinter or PyQt to adjust the image weights. 
def update(val):
    alpha = val / 100
    blended = cv2.addWeighted(img1, alpha, img2, 1 - alpha, 0)
    cv2.imshow('Blended Image', blended)

cv2.namedWindow('Blended Image')
cv2.createTrackbar('Blend', 'Blended Image', 0, 100, update)
update(0)
cv2.waitKey(0)
cv2.destroyAllWindows()

Blinking

  • Method: Display the two images alternately, blinking between them.
  • Advantage: Subtle differences show up immediately; effective when the two images are nearly identical.
  • Implementation: Use a timer to switch images at a fixed interval. 
while True:
    cv2.imshow('Comparison', img1)
    cv2.waitKey(500)  # 500ms 후 전환
    cv2.imshow('Comparison', img2)
    cv2.waitKey(500)

Swipe Line

  • Method: Partially cover one image with a mouse drag or slider to compare the same region of the two images.
  • Advantage: Accurate comparison is possible while directly checking the same region.
  • Implementation: Detect slider or mouse events and show one image cropped. 
def swipe(val):
    x = val
    swipe_img = img2.copy()
    swipe_img[:, :x] = img1[:, :x]
    cv2.imshow('Swipe Comparison', swipe_img)

cv2.namedWindow('Swipe Comparison')
cv2.createTrackbar('Swipe', 'Swipe Comparison', 0, img1.shape[1], swipe)
swipe(0)
cv2.waitKey(0)
cv2.destroyAllWindows()

Conclusion

  • For simple comparison, the Side-by-Side method is suitable.
  • For difference analysis, the Difference Overlay method is the most effective.
  • If intuitive interaction is needed, the Slider Blending or Swipe Line method is recommended.

See also

Favorite site

author Zer0 sitemap