Difference between the return values of PIL.Image.open and tf.image.decode_jpeg

Question

I use PIL.Image.open and tf.image.decode_jpeg to parse image files to arrays. But found that the pixel values in PIL.Image.open() are not the same as tf.image.decode_jpeg. Why this happens?

Thanks !

CodeOutput:

tf 100 100 [132 145 161]
pil 100 100 [134 147 164]

MyCode:

from __future__ import absolute_import
from __future__ import division
from __future__ import print_function

from datetime import datetime
import math
import time

import numpy as np
import tensorflow as tf

def decode_jpeg(image_file):
  from PIL import Image
  im = Image.open(image_file)
  data = np.array(im)
  return data

def tfimageread(filenames):
  filename_queue = tf.train.string_input_producer(filenames)
  reader = tf.WholeFileReader(name='image_reader')
  key, value = reader.read(filename_queue)
  uint8image = tf.image.decode_jpeg(value, channels=3)

  with tf.Session() as sess:
    coord = tf.train.Coordinator()
    threads = []
    for qr in tf.get_collection(tf.GraphKeys.QUEUE_RUNNERS):
      threads.extend(qr.create_threads(sess, coord=coord, daemon=True, start=True))
    image = sess.run(uint8image)
    coord.request_stop()
    coord.join(threads, stop_grace_period_secs=10)
    return image

if __name__ == '__main__':
  image_file = '。/o_1bchv9keu625336862107241874241888.jpg'
  image_tf = tfimageread([image_file])
  image_pil = decode_jpeg(image_file)
  i, j = 100, 100
  print ("tf %d %d %s" % (i,j,image_tf[i][j]))
  print ("pil %d %d %s" % (i,j,image_pil[i][j]))

A wild guess - are you sure both modules use the same coordinate system? — Błotosmętek, Jun 13 '17 at 07:39
Maybe Different decompression algorithm in PIL, tf, and opencv with jpeg images ? [Same question](https://stackoverflow.com/questions/31607731/opencv-vs-matlab-different-values-on-pixels-with-imread) | [Opencv issue](http://code.opencv.org/issues/4148) — xiusir, Jun 13 '17 at 07:49
Probably same problem as with [OpenCV and Pillow](https://stackoverflow.com/a/51121924/5407270). You need to make sure that PIL/Pillow uses same version of libjpeg as TensorFlow. — Andriy Makukha, Jul 01 '18 at 09:18

score 5 · Answer 1 · answered Feb 05 '20 at 15:43

A common cause of this problem is that tensorflow attempts to take shortcuts when decompressing jpegs. This offers a pretty large speedup for image reading, which can be the bottleneck for training certain CNNs, but does jitter the pixel values a bit.

Luckily, the developers have exposed options to turn off some of these efficiencies. In particular, check out the argument dct_method.

Try changing your call to tf.image.decode_jpeg to:

tf.image.decode_jpeg(value, channels=3, dct_method='INTEGER_ACCURATE')

You may also need to mess with fancy_upscaling depending on the sorts of images you're reading and other things going on in the underlying version of libjpeg that your software is using.

Difference between the return values of PIL.Image.open and tf.image.decode_jpeg

1 Answers1

Linked