pbiswell / pixeldifference

Calculate pixel differences between two images.

PixelDifference - Pixel differences between two images

Intro

This package is for comparing pixel differences between two images. It will check the same pixel positions on both images. This can be used to compare how different an image is to the original after compression or editing, for example.

Due to some functions being relatively slow on large images, I will recode them into C in a future update.

Please note that I am not educated in colour science. If there are any mistakes in implementations or calculations, please let me know and I will fix them as soon as possible. Thank you.

Requirements

You need python3, pillow, and numpy to use this package.

pip install pillow

pip install numpy

Installation

pip install pixeldifference

Usage

# Import packages
from pixeldifference import PixelDifference
from PIL import Image

# Load two images from your device
# You need to change the image paths to your images
imageOne = Image.open('/path/to/imageOne.jpg', 'r')
imageTwo = Image.open('/path/to/imageTwo.jpg', 'r')

# Initialise
pd = PixelDifference(imageOne, imageTwo)

# The total number of pixels checked
totalpixels = pd.total

# The percentage of different pixels compared to the total checked area
percentdifferent = pd.percent

# The total number of different pixels in the checked area
pixelsdifferent = pd.pixels

Advanced Settings

Average Colour Distance

The following functions calculate colour difference between all selected pixels in two images, and provide averages.

Simple RGB Euclidean Distance

This uses the following formula to calculate the colour distance between all of the selected pixels.

Link: https://en.wikipedia.org/wiki/Color_difference#sRGB

ed = PixelDifference(imageOne, imageTwo).EuclideanDistance()

# Average value
ed.average
# Minimum value
ed.min
# Maximum value
ed.max
# Median value
ed.median
# Time taken
ed.elapsed

More Accurate RGB Euclidean Distance

This uses the following formula to calculate the colour distance between all of the selected pixels.

Link: https://en.wikipedia.org/wiki/Color_difference#sRGB

ed2 = PixelDifference(imageOne, imageTwo).EuclideanDistance2()

# Average value
ed2.average
# Minimum value
ed2.min
# Maximum value
ed2.max
# Median value
ed2.median
# Time taken
ed2.elapsed

Redmean

This uses the following formula to calculate the colour distance between all of the selected pixels.

Link: https://en.wikipedia.org/wiki/Color_difference#sRGB

redmean = PixelDifference(imageOne, imageTwo).Redmean()

# Average value
redmean.average
# Minimum value
redmean.min
# Maximum value
redmean.max
# Median value
redmean.median
# Time taken
redmean.elapsed

CIE76 - LAB Colour Distance

This converts the image from RGB to LAB, and uses the following formula to calculate the average CIE76 distance between all of the selected pixels in the CIELAB colourspace.

Link: https://en.wikipedia.org/wiki/Color_difference#CIE76

cie76 = PixelDifference(imageOne, imageTwo).CIE76()

# Average value
cie76.average
# Minimum value
cie76.min
# Maximum value
cie76.max
# Median value
cie76.median
# Time taken
cie76.elapsed

Images of Different Size

Update 1.0.1: Will now automatically crop images to an area which exists in both images, starting at the top left corner.

Image File Sizes

You can compare the file sizes of the images. File sizes are returned in bytes.

pd = PixelDifference(imageOne, imageTwo)

# File size of image 1 in bytes
filesize1 = pd.filesize[0]

# File size of image 2 in bytes
filesize2 = pd.filesize[1]

# File size difference in bytes, of image 2 compared to image 1
# If image 2 is smaller, the result is negative
diff = pd.filesizedifference

Time Elapsed

You can see how long it took to calculate with .elapsed

pd = PixelDifference(imageOne, imageTwo)

# Elapsed time to calculate differences
pd.elapsed

Features

• Compare two images for differences, pixel by pixel
• Use different sized images
• Image difference in pixels, and percentage of total area

To Do

• Improve file size comparison
• Improve Readme
• Command line arguments
• More tests
• Handle images from string paths
• Performance tests/improvements
• Allow for more than 2 images
• More comparisons in LAB
• Code slow operations in C
• Get colour of specific pixels
• Transparency

Changes

-v1.0.1

- Added file sizes and comparison
Usage:
pd.filesize[0]
pd.filesize[1]
pd.filesizedifference

- Added time taken
Usage:
pd.elapsed

- Added overall average percentage difference
Usage:
print(wa.overalldifference)

- Added CIE76 distance
- Added RGB simple euclidean distance
- Added More accurate euclidean distance
- Added Redmean

-v1.0.0 (June 12th, 2021)

Initial release