# This sample shows the an efficient way of doing image processing
# over Blender's images using Python.

import bpy
import numpy as np


input_image_name = "Image"
output_image_name = "NewImage"

# Retrieve input image.
input_image = bpy.data.images[input_image_name]
w, h = input_image.size

# Allocate a numpy array to manipulate pixel data.
pixel_data = np.zeros((w, h, 4), 'f')

# Fast copy of pixel data from bpy.data to numpy array.
input_image.pixels.foreach_get(pixel_data.ravel())

# Do whatever image processing you want using numpy here:
# Example 1: Inverse red green and blue channels.
pixel_data[:,:,:3] = 1.0 - pixel_data[:,:,:3]
# Example 2: Change gamma on the red channel.
pixel_data[:,:,0] = np.power(pixel_data[:,:,0], 1.5)

# Create output image.
if output_image_name in bpy.data.images:
    output_image = bpy.data.images[output_image_name]
else:
    output_image = bpy.data.images.new(output_image_name, width=w, height=h)

# Copy of pixel data from numpy array back to the output image.
output_image.pixels.foreach_set(pixel_data.ravel())
output_image.update()