Sorting digital photos according to the real creation time using EXIF tags.

Recently I was on a holiday with my friends. Since there were a lot thrilling places to see, almost everybody took a camera. Finally after 8 days of intensive hiking we had over 900 photos taken and we wanted to make a chronological album. 

The first problem that occurred to us is that the global image namespace was not continuous (5 different cameras / cellphones with the ability to take photos), and sorting images by their generated names did not solve the it. 

Next we tried sorting the images by the file modification date - this was a reasonable approach, but there a few photos which were rotated after fetching them from the memory card causing their modification date to be inadequate for the mentioned problem. Furthermore 2 cameras were configured with an incorrect date / time value. 

A few years ago I was implementing an EXIF tag viewer for images, and I remembered that there are tags that store the original image creation time (and quite a few other parameters). I made some research and located a library with python bindings that handles these tags well: pyexiv2. 

You may install it using apt-get:

~ $ apt-get install python-pyexiv2

Next I implemented a simple script updates file names so they can be sorted lexicographically  according to the real creation time, have a look at the following code:

import pyexiv2
import sys
import os
import shutil
import datetime
from dateutil.relativedelta import relativedelta

class ExifSort():
    @classmethod
    def _test_extenstion(cls, filename):
        return filename.split(".")[-1].lower() in ['jpg','jpeg','png']
    @classmethod
    def sort_images(cls, path, prefix=None):
        files = []
        for file in os.listdir(path):
            if not cls._test_extenstion(file):
                continue
            metadata = pyexiv2.ImageMetadata(os.path.join(path, file))
            metadata.read()
            try:
                d = metadata['Exif.Photo.DateTimeOriginal'].value
                if metadata['Exif.Image.Make'].value == 'Canon':
                    d+=datetime.timedelta(days=1)
                elif metadata['Exif.Image.Make'].value == 'FUJIFILM':
                    d += relativedelta(years=1)    
                    d -= datetime.timedelta(hours=39)
            
                files.append((file, d, )) 
            except:
                print file
                continue

        files.sort(key=lambda x: x[1])
        mag = len(str(len(files))) 
        counter = 1
        tmp_prefix = "tmp_"
        for data in files:
            shutil.move(os.path.join(path, data[0]), os.path.join(path, tmp_prefix + data[0]))
        for data in files:
            ext = data[0].split(".")[-1]
            shutil.move(os.path.join(path, tmp_prefix + data[0]),\
                    os.path.join(path, "%s%s.%s" % (prefix or "",\
                    str(counter).rjust(mag,'0'), ext)))
            counter += 1


if __name__ == '__main__':
    try:
        ExifSort.sort_images(sys.argv[1])
    except:
        print "Usage: %s [directory]" % sys.argv[0]
        sys.exit(1)

This script iterates all files in a specific directory ( sys.argv[1] ), filters jpg/png files, reads their exif date and stores pairs: filename, creation datetime in an array. Finally the array is sorted using the real creation datetime, and the images are renamed (with or withour a specific prefix).   

Lines 22-26 are a specific exception - we had 2 cameras with an incorrect date, so we have to correct the read value so it may be compared with images from other cameras (evaluated empirically) . 

~KR