How to generate mosaic painting in python? Method to generate mosaics (detailed code explanation)

The content of this article is to introduce how to generate mosaic paintings in python? Method to generate mosaics (detailed code explanation). It has certain reference value. Friends in need can refer to it. I hope it will be helpful to you.

Mosaic painting is a big picture made up of small pictures. Our renderings, zoom in to see the details. Each piece is an independent picture. When put together to form a big picture, it feels like A painting made of mosaics, so it is called mosaic painting. I saw some mosaics on the Internet and thought they were cool, so I used Python to convert an original image into a mosaic.

Our rendering is like this:

The original picture is like this:

The specific idea of ​​implementation is as follows:

Step one: First collect a set of pictures, which will be used as small square pictures in the big picture. The more pictures there are, the closer the color of the final generated picture is.

Step 2: Divide the picture to be converted into small square pictures, like the following

Step 3: For each small square Grid picture, replace it with the closest picture in the picture set. After all the small squares are replaced, our final mosaic is generated.

Sounds simple?

Let’s take a look at the specific implementation steps. Here are some core codes.

Our picture collection is stored in the images directory. The following code loads all the pictures in the directory and scales them to a unified size

import re
import os
import cv2
import numpy as np
from tqdm import tqdm
IMG_DIR = "images"
def load_all_images(tile_row, tile_col):
 img_dir = IMG_DIR
 filenames = os.listdir(img_dir)
 result = []
 print(len(filenames))
 for filename in tqdm(filenames):
 if not re.search(".jpg", filename, re.I):
 continue
 try:
 filepath = os.path.join(img_dir, filename)
 im = cv2.imread(filepath)
 row = im.shape[0]
 col = im.shape[1]
 im = resize(im, tile_row, tile_col)
 result.append(np.array(im))
 except Exception as e:
 msg = "error with {} - {}".format(filepath, str(e))
 print(msg)
 return np.array(result, dtype=np.uint8)

The parameter of the load_all_images function here is the unified size, tile_row and tile_col correspond to height and width respectively.

The following code divides the picture to be converted

img = cv2.imread(infile)
tile_row, tile_col = get_tile_row_col(img.shape)
for row in range(0, img_shape[0], tile_row):
 for col in range(0, img_shape[1], tile_col):
 roi = img[row:row+tile_row,col:col+tile_col,:]

We divide the picture to be converted into small squares. tile_row and tile_col are the height and width of the small squares. Roi storage Get the image data in the small square.

The following is a function to calculate the similarity of two pictures

from scipy.spatial.distance import euclidean
def img_distance(im1, im2):
 if im1.shape != im2.shape:
 msg = "shapes are different {} {}".format(im1.shape, im2.shape)
 raise Exception(msg)
 array1 = im1.flatten()
 array2 = im2.flatten()
 dist = euclidean(array1, array2)
 return dist

im1 and im2 are the data of two pictures. The picture data is a three-dimensional numpy array. Here we convert the three-dimensional array into one-dimensional After the array, compare the Euclidean distance between the two. To find the most similar picture, you only need to traverse all the pictures in the picture set, find the picture with the shortest distance, and replace the small squares in the original picture.

Let’s take a look at the final effect:

The partial details in the enlarged picture are as follows:

If you are not satisfied with the picture quality and want a more refined picture quality, you can consider dividing the picture into smaller squares during segmentation, but this will also increase the running time of the program.

The above is the detailed content of How to generate mosaic painting in python? Method to generate mosaics (detailed code explanation). For more information, please follow other related articles on the PHP Chinese website!