Image Palette

Extract the dominant colours from any image and either add the palette to the source image, generate it as a separate file or get it in a JSON file.

This program uses k-mean clustering (naively implemented by myself) to determine the most prevalent colours. It's quite resources consuming but with a few precautions, it can be made faster.

Look down below for sample images or check my Instagram profile to see more of my work! If you want to replicate this effect in your own photos, make sure to scroll a little bit further for the instructions.

Output

Input image

Generated colour palette

Palette integrated into original image

Output as seen in the console

JSON file with the colour palette

Filename starry-night-json-palette.json

Content:

{
  "rgb": [
    [168, 177, 152],
    [110, 134, 156],
    [23, 28, 33],
    [69, 93, 137],
    [47, 60, 89 ]
  ],
  "hsv": [
    [81, 14, 69],
    [208, 29, 61],
    [210, 30, 12],
    [218, 49, 53],
    [221, 47, 34]
  ],
  "hex": [
    "#A8B198",
    "#6E869C",
    "#171C21",
    "#455D89",
    "#2F3C59"
  ]
}

How to use

• Install Python3 on your PC
• Clone the repository or download the release
• Navigate to the folder containing the script
• Install the required Python modules with the command python3 -m pip install -r requirements.txt
• View a list of all the available arguments python3 imagepalette.py -h

Image resizing

By using the command --resize (or -r) the image will be resized in order to speed up the extraction of the color. The output image will not be affected and will be the same size as the original one.

Arguments

Command	Description	Optional	Defaults	Type
-i --input	Source image path	✗	none	string
-o --output	Custom output folder	✓	output/	string
-c --colors	Number of extracted colors	✓	5	int
-r --resize	Resize the image for internal use	✓ recommended (see below)	none	none
--console	Log to console	✓	False	none
--palette	Create an image containing the palette	✓ one of this group must be selected	none	none
--print	Print the palette in the console	✓ one of this group must be selected	none	none
--json	Create a JSON file containing the palette	✓ one of this group must be selected	none	none
--incorporated	Incorporate the palette inside the original image	✓ but one of this group must be selected	none	none
--scl	Relative of the original image in the output image. (valid if used in the incorporated mode)	✓	0.9	float
--color-width-scl	Width of the color rectangle relative to the size of the palette (valid if used in the incorporated mode)	✓	0.9	float
--color-height-scl	Ratio of the palette to the original image (valid if used in the incorporated mode)	✓	0.9	float
--position	Position of the color legend (valid if used in the incorporated mode)	✓	r	{l, r, t, b}
--color	Background color of the image (valid if used in the incorporated mode)	✓	220 220 220	int int int
--outline	Outline color of the palette (valid if used in the incorporated mode)	✓	none	int int int
--outline-width	Width of the outline of the palette (valid if used in the incorporated mode)	✓	1	int
--seed	Seed for the random number generator	✓	none	int
--min-color-distance	Minimum distance between colors (valid if used in the incorporated mode)	✓	35	float
--max-iterations	Maximum number of iterations for the color extraction without change in the objective funciton	✓	5	int

Resize argument

By setting this flag, the image will be resized before being processed. This won't affect the final result size and will speed up the process. The only downside is that there could be a very little loss of colour, but will be likely not visible.

Examples

• Generate a 4 colour palette of the image "image-1.png" and save it as a new image file: python3 imagepalette.py -i image-1.png -c 4 --palette
• Generate a 4 colour palette of the image "image-1.png" and print it in the console: python3 imagepalette.py -i image-1.png -c 4 --print
• Generate a 8 color palette of the image "image-1.png" and save it as a JSON file: python3 imagepalette.py -i image-1.png -c 8 --json
• Generate a 5 colour palette of the image "image-1.png" and incorporate it in the source image: python3 imagepalette.py -i image-1.png -c 5 --incorporated
  • The arguments --palette-width, --palette-height, --scl, --position, --color, --outline, --no-outline can be used to further customize the output image
• Generate a 5 colours palette of the image "image-1.png" and incorporate it on the left side of the source image with a purple background and a gold outline 5 pixels wide: python3 -i image-1.png --position l --incorporated --color 128 0 128 --outline 255 215 0 --outline-width 5

Batch converting

I have included a script called batch-convert.py that will automatically create integrated palettes for all the `.jpg`` files placed in the same folder.

License

This project is distributed under the MIT License. See LICENSE.md for more information.