Automating Image clean up with Python

If you have been reading this blog, you'll know I collect sports cards. It's fun to share what I have with other collectors (by posting scans on sportscollectors.net, facebook, etc.).

A few years ago I bought a Brother MFC-9130CW all in one printer/scanner that I use to do my scanning. I usually set it to scan documents in legal format so I can fit 9 cards on a scan (3 rows by 3 columns). And since I want to do this most efficiently, I generally save them as a 200dpi pdf file with multiple pages.

The requirements:

1. We will have a multipage pdf, with each page containing a 3x3 grid of equal sized card images. I will use this sample file for this exercise: sample_cards_file.pdf (3.02 mb)
   
   
2. We will need each image cropped to show the full card (we can have some extra space around it) and saved to it's own jpeg file. An example is show below:
   
   
    
3. I will supply a directory path containing the outputted images. Since I will put them in my inventory database, I'll need to supply the starting number for the image names and have each subsequent image increment by one.
   
   
   
   

The process: 

I like using Python for automating things, as it seems to have libraries for most things I want to do.  So I figured it would be a good candidate for this project.

1. The first thing we will want to be able to define some global variables
   
   

   #paths
   source_file = '/tmp/sample_cards_file.pdf'
   out_dir = '/tmp/process/'
   
   # page setup
   rows = 3
   cols = 3
   
   # defines where the last run of this left off on (first item will be 1.jpg if 0)
   starting_count = 0
   
   #spacing offsets
   top_offest = 0
   left_offset = 260
   bottom_offset = 0
   right_offset = 0
   spacer = 40
   vert_spacer=0

2. We will need to do is convert the each page of the pdf to a jpeg file.
   
   Python has a library called pdf2image to do this:
   
   

   pip install pdf2image
   pip install poppler
   
   # I used this syntax instead do of pip when running this in a jupyter notebook
   # conda install -c conda-forge pdf2image
   # conda install -c conda-forge poppler
   
   from pdf2image import convert_from_path
   
   # function that converts multipage pdf to individual
   # jpeg images. Function returns list of image paths.
   def convert_pdf_to_jpegs(pdf_path, out_dir):
       file_paths=[]
       pages = convert_from_path(pdf_path, 500)
       page_count = 1
       for page in pages:
           image_path="{}temp_page_{}.jpg".format(out_dir, str(page_count))
           #add to the list
           file_paths.append(image_path)
           #save converted file
           page.save(image_path, 'JPEG')
           page_count += 1
       return file_paths

3. We will have a function that breaks up an image with our 9 cards and save each individually to a specified directory. We will need the ability to seed the first image name. The function can use the spacing offset variables
   
   

   from PIL import Image
   
   def split_images_from_page(image_path, out_dir, row_count, col_count, start_number):
       # opens the image file 
       Im = Image.open(image_path)
   
       # calculates height and width of image
       full_width = Im.width
       full_height = Im.height
       image_height = int((full_height - vert_spacer - top_offest - bottom_offset) / rows)
       image_width = int((full_width - spacer - left_offset - right_offset) / cols)
       
       image_count = 0
       row_current=1
   
       #iterates through rows and columns
       while row_current <= row_count:
           col_current = 1
           while col_current <= col_count:
   
               # calculates the coordinates on the image to crop            
               croppedIm = Im.crop((left_offset + ((col_current - 1) * image_width) + spacer, top_offest + ((row_current - 1) * image_height), min(left_offset + (col_current * image_width) + ((col_current - 1) * spacer), Im.width), top_offest + (row_current * image_height) + vert_spacer))
   
               # if you wanted to resize the image to 300 width and 420 height 
               # croppedIm = croppedIm.resize((300, 420))
               
               # saves the image to specified directory
               croppedIm.save("{}/{}.jpg".format(out_dir, start_number+image_count))
          
               col_current += 1
               image_count += 1        
           row_current+=1

4. Calling the functions:
   
   

   import os
   
   page_count = 1
   
   # convert pdf to jpeg file for each page
   file_paths = convert_pdf_to_jpegs(source_file, out_dir)
   
   # split each page into images
   for file_path in file_paths:
       number_start = starting_count + (page_count-1) * (rows * cols) + 1 
       split_images_from_page(file_path, out_dir, rows, cols, number_start)
       page_count += 1
   
   # clean up delete jpeg files for each page
   for file_path in file_paths:
       os.remove(file_path)

Here is my ipython notebook with the code detailed above: 

pdf_scan_process.ipynb