Pass PIL Image Objects to ColorThief and Ignore White/Black Pixels

[gsornsen]

Thanks for your work on Color Thief! It's a really excellent tool.

I'm working on a Linux project to sync the dominant color of my primary display to my Hue Lights. Unfortunately, Hue Sync only works on Windows/Mac.

I'd rather not be I/O bound having to read/write to a disc to get the dominant color and would like to use your tool as a dependency. This pull request reflects the necessary changes to make that happen.

Here is a code snippet using the modified ColorThief class:

#!/usr/bin/env python3

import pyscreenshot as ImageGrab
from displays import Monitor # Local import
from PIL import Image
from colorthief import ColorThief

# Get screen information from primary display
monitor = Monitor()
# Use pyscreenshot to grab a Linux screenshot using a native library
screen_grab = ImageGrab.grab(bbox=(monitor.x, monitor.y, monitor.width, monitor.height))
# Convert the image to a compatible PIL Image object
img = Image.frombytes('RGB', screen_grab.size, screen_grab.tobytes())
# Feed the PIL Image object to ColorThief
color_thief = ColorThief(img, True)
dominant_color = color_thief.get_color(quality=10000)
print(dominant_color)

[14ROVI]

Hello,
I also started with this problem and i agree that it would be nice to be able to use Pillow images directly with it. However, a work around that works with the current:

with io.BytesIO() as file_object:
    pillow_image.save(file_object, "PNG")
    cf = ColorThief(file_object)
    colours = cf.get_color()

As you can see, this is quite elegant in its own sense too and it doesn't write to disc.

Hope that helps you!

just realised that this is from 2019 lmao

[ShubhamKhandare]

Hi,
I have directly passed screenshot to ColorThief and its seems to be working fine.

# Use pyscreenshot to grab a Linux screenshot using a native library
screen_grab = ImageGrab.grab(bbox=(monitor.x, monitor.y, monitor.width, monitor.height))
# Feed the PIL Image object to ColorThief
color_thief = ColorThief(screen_grab, True)