This Python package implements blend modes for images.
Blending through blend modes allows to mix images in a variety of ways.
This package currently supports the following blend modes (name of the
respective functions in the package in
- Soft Light (
- Lighten Only (
- Dodge (
- Addition (
- Darken Only (
- Multiply (
- Hard Light (
- Difference (
- Subtract (
- Grain Extract (known from GIMP,
- Grain Merge (known from GIMP,
- Divide (
- Overlay (
- Normal (
The intensity of blending can be controlled by means of an opacity parameter that is passed into the functions. See Usage for more information.
The Blend Modes package is optimized for speed. It takes advantage of vectorization through Numpy. Further speedup can be achieved when implementing the package in Cython. However, Cython implementation is not part of this package.
The blend mode functions take image data expressed as arrays as an input. These image data are usually obtained through functions from image processing packages. Two popular image processing packages in Python are PIL or its fork Pillow and OpenCV. The examples in this chapter show how to blend images using these packages.
Input and Output Formats¶
A typical blend mode operation is called like this:
blended_img = soft_light(bg_img, fg_img, opacity)
The blend mode functions expect
Numpy float arrays in the
format [pixels in dimension 1,pixels in dimension 2,4] as an
input. Both images needs to have the same size, so the pixels in
dimension 1 must be the same for
applies to the pixels in dimension 2. Thus, a valid shape of the
arrays would be
bg_img.shape == (640,320,4) and
fg_img.shape == (640,320,4).
The order of the channels in the third dimension should be R, G, B, A, where A is the alpha channel. All values should be floats in the range 0.0 <= value <= 255.0.
The blend mode functions return arrays in the same format as the input format.
The following examples show how to use the Blend Modes package in typical applications.
The examples are structured in three parts:
- Load background and foreground image. The foreground image is to be blended onto the background image.
- Use the Blend Modes package to blend the two images via the “soft light” blend mode. The package supports multiple blend modes. See the Description for a full list.
- Display the blended image.
from PIL import Image import numpy from blend_modes import soft_light # Import background image background_img_raw = Image.open('background.png') # RGBA image background_img = numpy.array(background_img_raw) # Inputs to blend_modes need to be numpy arrays. background_img_float = background_img.astype(float) # Inputs to blend_modes need to be floats. # Import foreground image foreground_img_raw = Image.open('foreground.png') # RGBA image foreground_img = numpy.array(foreground_img_raw) # Inputs to blend_modes need to be numpy arrays. foreground_img_float = foreground_img.astype(float) # Inputs to blend_modes need to be floats. # Blend images opacity = 0.7 # The opacity of the foreground that is blended onto the background is 70 %. blended_img_float = soft_light(background_img_float, foreground_img_float, opacity) # Convert blended image back into PIL image blended_img = numpy.uint8(blended_img_float) # Image needs to be converted back to uint8 type for PIL handling. blended_img_raw = Image.fromarray(blended_img) # Note that alpha channels are displayed in black by PIL by default. # This behavior is difficult to change (although possible). # If you have alpha channels in your images, then you should give # OpenCV a try. # Display blended image blended_img_raw.show()
The following example shows how to use the Blend Modes package with OpenCV.
import cv2 # import OpenCV import numpy from blend_modes import soft_light # Import background image background_img_float = cv2.imread('background.png',-1).astype(float) # Import foreground image foreground_img_float = cv2.imread('foreground.png',-1).astype(float) # Blend images opacity = 0.7 # The opacity of the foreground that is blended onto the background is 70 %. blended_img_float = soft_light(background_img_float, foreground_img_float, opacity) # Display blended image blended_img_uint8 = blended_img_float.astype(numpy.uint8) # Convert image to OpenCV native display format cv2.imshow('window', blended_img_uint8) cv2.waitKey() # Press a key to close window with the image.
The Blend Modes package can be installed through pip:
pip install blend_modes
The Blend Modes package needs Numpy to function correctly. For loading images the following packages have been successfully used:
I am happy about any contribution or feedback. Please let me know about your comments via the Issues tab on GitHub.