Search a page for a representative image (a big, square one). Inspired by the code Reddit uses to pick images to put next to headlines

ThumbnailGetter.py

import io
import sys
import urllib

import bs4
import requests
from PIL import Image

def get_image_list(url):
    """Given the url of an HTML document, return a list of urls of all images
    on the page.
    """
    response = requests.get(url)
    soup = bs4.BeautifulSoup(response.content, "html.parser")
    image_tags = soup.findAll('img')

    image_urls = []
    for image_tag in image_tags:
        image_url = image_tag.get('src')

        # image_url might be a relative url, urljoin will make a full url if necessary
        full_url = urllib.parse.urljoin(url, image_url)

        if full_url[-4:] != '.svg':  # PIL can't handle svgs, so skip
            image_urls.append(full_url)

    return image_urls

def image_dimensions(image):
    width = image.size[0]
    height = image.size[1]
    return width, height, width*height

def is_square(width, height):
    """Determine if an image is close to square shaped"""
    return 0.5 < width/height < 2

def is_large(area):
    """Determine if an image is largish"""
    return area > 5000

def select_best_image(image_urls):
    biggest_image = None
    biggest_size = 0

    for url in image_urls:
        response = requests.get(url)
        im = Image.open(io.BytesIO(response.content))
        width, height, area = image_dimensions(im)

        if is_square(width, height) and is_large(area):
            if area > biggest_size:
                biggest_image = url
                biggest_size = area

    return biggest_image

if __name__ == '__main__':
    url = sys.argv[1]
    images = get_image_list(url)
    print(select_best_image(images))