tags: image blend pillow numpy

This is a simple script used to generate the background of this page by blending the horizontal edges
of an image. It is rough and ready, written primarily just for this particular image, but illustrates
some basic Pillow and Numpy. I've left the comments in with my rough workings of the maths.

(For those interested, the background image was taken [from here in pickpik.com](https://www.pickpik.com/brown-tree-purple-sky-golden-hour-purple-sunset-91149) — it was found by specifying Creative Commons in
the Google image search tools, but I could find no license on the page, so I presume it's ok
to use like this on a personal wiki.)

```py
#!/usr/bin/env python3
from PIL import Image
import numpy as np

image = Image.open("pt_bg_001.png")
data = np.asarray(image)
height, width, _ = data.shape

blend_width = 64
bw = blend_width
bw2 = int(blend_width//2)
blend_width = 2*bw2

# output width is input width - 
output_width = width - blend_width
ow = output_width
ow2 = ow - bw2
odata = np.zeros((height,output_width,3),dtype = data.dtype)

for i in range(output_width):
  for j in range(height):
    if i < bw2:
      for k in range(3):
        # pixel 0 is 50/50
        # pixel bw2 is 100% 

        # when i==bw2, this==1.0
        # when i==0, this==0.5
        pc = 0.5*(1+(i/bw2)) 
        pd = 1-pc

        b = data[j,i+bw2,k]
        a = data[j,width-bw2+i,k]
        c = int(a*pd+b*pc)
        odata[j,i,k] = c
    elif i > output_width - bw2:
      for k in range(3):
        # pixel 0 is 50/50
        # pixel bw2 is 100% 

        ii = i - ow2
        # so ii now runs from 0 to bw2
        a = data[j,i,k]
        b = data[j,i-ow2,k]
        pc = 0.5*(ii/bw2) 
        pd = 1-pc

        c = int(a*pd+b*pc)
        odata[j,i,k] = c
    else:
      odata[j,i] = data[j,i+bw2]

image2 = Image.fromarray(odata)
image2.save("out.png")

image2.save("pt2_bg1.jpg")

```