auto-barfilter

#!/usr/bin/env python3
import argparse

helpmsg = "Infers and creates a bar filter to change edge color"
parser = argparse.ArgumentParser(
    description=helpmsg)
parser.add_argument('-b', '--buffer', type=int, nargs=1, default=[0],
        help="Number of pixels to add to border size")
parser.add_argument('-t', '--tolerance', type=float, default=1.0)
parser.add_argument('-v', '--verbose', action='store_true')
parser.add_argument('-o', '--output', required=True, help="Name of output file")
parser.add_argument('filename', help="Images to concatenate")

args = parser.parse_args()

from PIL import Image
import numpy as np
from numpy.linalg import norm

im = Image.open(args.filename)
im = np.asarray(im)

# starting from the left, count the number of pixels before the change in
# avg/std is greater than the tolerance
height, width, nchannels = im.shape
col = im[:,0,:]
prev = np.array([col.mean(axis=0), col.std(axis=0)])
for w in range(1, width):
    col = im[:,w,:]
    mean = col.mean(axis=0)
    std = col.std(axis=0)

    # find the first significantly different column
    this = np.array([mean, std])
    delta = norm(this - prev)

    if args.verbose:
        print('prev:')
        print(prev)
        print('this:')
        print(this)
        print('delta:')
        print(delta)

    if delta > args.tolerance:
        break

    if args.verbose:
        print('-'*76)
        
    prev = this

border_width = w-1 + args.buffer[0]
border_shape = height, border_width, 4

middle_region_shape = height, width - border_width*2, 4

# make an image with white at the edges, transparent in the middle
white = np.uint8([255, 255, 255, 255])
transparent = np.uint8([0, 0, 0, 255])
border = np.ones(border_shape, dtype=np.uint8) * white
middle = np.zeros(middle_region_shape, dtype=np.uint8) * transparent
output = np.concatenate([border, middle, border], axis=1)

# do the actual concatenation along the specified dimension
output = Image.fromarray(output)
output.save(args.output)