Re: [Gimp-user] Is there a single bit that can accidentally flip an image to negative? (not Gimp specific)

[Ofnuts <ofnuts laposte net>]

On 07/28/2013 03:28 AM, Ofnuts wrote:
> On 07/28/2013 02:41 AM, Jay Smith wrote:
> > Hi,
> >
> > I am sorry that this query is not Gimp specific, but since I use Gimp 
> > to create the images and I don't subscribe to any "images" list, I 
> > hope it will be allowed here.  I won't make a habit of it.
> >
> > From time to time I notice an image in our library or on our website 
> > has "become" negative.  This is very unsettling as I have used such 
> > images on our website for years and then all of a sudden I notice in 
> > the middle of a web page an image that is in negative.
> >
> > Yes, it is theoretically possible that I simply never noticed this, 
> > but I have been finding one every few months now for three or four 
> > years.  I know everybody says "I did not do it", but these are images 
> > which have not intentionally been manipulated, processed, or uploaded 
> > to the hosting service for for years. The differences in image 
> > appearance are so dramatic that I just can't believe that they have 
> > been like that all along and I just did not see them.
> >
> > Also, the images don't seem to suffer any other type of damage. They 
> > just become negative.
> >
> > In the most recent example case (see links below), only the file 
> > residing on the hosting service has "become" negative.  The original 
> > JPEG that I uploaded is still correct!  (However, over the last 
> > couple years, I have found an occasional "negative-ized" original 
> > JPEG on our server and also a few "negative-ized" TIFFs on our 
> > server.  It seems that the alteration to negative can happen anywhere 
> > along the line.
> >
> > Since the source JPEG version residing on my server is still okay, 
> > the damage happened either during transfer to the hosting service or 
> > while on the hosting service.
> >
> >
> > Here is an example (the two are images of slightly different objects, 
> > but the overall appearance is supposed to be extremely similar -- the 
> > person who can determine the difference is a keen-eyed philatelist):
> >
> > CORRECT:
> > http://jsa.viewimage.net/jsa/web/Lists/Sweden/Seals/FlagDay/flag-day-1936-perf-11_mint_149571_r_l.jpg 
> >
> >
> > NEGATIVE:
> > http://jsa.viewimage.net/jsa/web/Lists/Sweden/Seals/FlagDay/flag-day-1936-perf-11-5-x-11_mint_128795_r_l.jpg 
> >
> >
> >
> > My process: All on Linux.  After scanning to TIFF, the images are 
> > edited in Gimp as TIFFs. Then they are processed using a script that 
> > runs various ImageMagick actions to create various sizes (copies) as 
> > JPEGs. The source TIFF is preserved unchanged (and the original TIFF 
> > for the negative example above is CORRECT!).  The JPEGs are then 
> > uploaded to a hosting service (in this case all the 'original' JPEGs 
> > still on my server are CORRECT!!).
> >
> > So... The question is.....  is there a bit in an image file that can 
> > be hit by a "stray neutrino" (or whatever happens) that can cause the 
> > image file to "become" negative".  If yes for JPEG, is it also 
> > possible for TIFF?
> >
> > Or, it seems more likely to me, would such a change from positive to 
> > negative require a very large number of changes to the image file?
> >
> > If the type of occasional damage I am observing in a "small" library 
> > of about 100,000 image files is happening everywhere, to everybody, 
> > to all types of computer files, we are in deep do-do in the long 
> > term.  As they say.... "Houston, we have a problem."
> Actually it's not a negative... and the changes would be on more than 
> one bit since the bad image file is about 50% bigger than the good one.
>
> But the images aren't even the same size (314x431 vs 312x426), and if 
> you look closely in the corners the perforations aren't the same (you 
> can invert the colors of the bad ones to make it more visible). So 
> these are two images of different stamps.
>
> The 'bad' image has a color profile, and this one may be incorrect.

Hmmm. Answered without reading the end of the post... So these are 
indeed different subjects... But if you want anyone to determine what 
happens when an image is corrupted you have better provide the 
before/after versions of the same image.

A bad hard disk or file system problems can cause A JPEG to become 
corrupt but the look is different.

Image hosts often have a cavalier attitude wrt to the images deposited 
on them (arbitrary format changes, quality reduction...) but I have 
never heard of any adding a color profile.

If that can help the profile seems to be added by some Hewlett-Packard 
software, but I've never seen any HP software running on Linux and your 
whole process is on Linux.