I suspect gtk uses for decoding of jpeg images the functions of the os, which might have changed between the two version. Encoding of images in jpeg is lossy. And so might the decoding not be exact.
http://en.wikipedia.org/wiki/Jpeg#Required_precision.
Have you tried an bmp file wich is lossless?
By the way using python2 and python-imaging with:
Code:
#!/usr/bin/env python2
# -*- coding: utf-8 -*-
import Image
histo=Image.open("tp1.JPG").convert("L").histogram()
for i in range(222,256):
print i, histo[i]
# How long is your program?
I'll get:
Code:
222 14
223 8
224 8
225 10
226 15
227 10
228 9
229 12
230 12
231 12
232 14
233 7
234 22
235 12
236 14
237 14
238 17
239 19
240 23
241 25
242 34
243 35
244 20
245 32
246 40
247 30
248 35
249 54
250 73
251 78
252 177
253 292
254 536
255 12375