Main Page | Report this Page
 
   
Science Forum Index  »  Compression Forum  »  Corrupt JPEG data
Page 1 of 1    
Author Message
mathieu
Posted: Thu Apr 17, 2008 3:36 am
Guest
Hi there,

I just found yet another broken JPEG file:

http://gdcm.sourceforge.net/thingies/GENESIS_SIGNA-JPEG-CorruptFrag.jpg

jpeginfo gives me:

$ ./bin/jpegdump16 -verbose /tmp/GENESIS_SIGNA-JPEG-
CorruptFrag.jpg
Independent JPEG Group's JPEGDUMP, version 6b 27-Mar-1998
Copyright (C) 1998, Thomas G. Lane
Start of Image
Define Huffman Table 0x00
Start Of Frame 0xc3: width=256, height=256, components=1
Component 0: 1hx1v q=0
Start Of Scan: 1 components
Component 0: dc=0 ac=0
Ss=1, Se=0, Ah=0, Al=0
monochrome
lossless (sequential)
Huffman

trying to decompress the image I am getting:

$ ./bin/djpeg16 -outfile dummy.pnm /tmp/GENESIS_SIGNA-JPEG-
CorruptFrag.jpg
Corrupt JPEG data: 121 extraneous bytes before marker 0xd9

Thanks for any info

-Mathieu
Thomas Richter
Posted: Thu Apr 17, 2008 2:16 pm
Guest
mathieu wrote:

Quote:
I just found yet another broken JPEG file:

http://gdcm.sourceforge.net/thingies/GENESIS_SIGNA-JPEG-CorruptFrag.jpg

jpeginfo gives me:

$ ./bin/jpegdump16 -verbose /tmp/GENESIS_SIGNA-JPEG-
CorruptFrag.jpg
Independent JPEG Group's JPEGDUMP, version 6b 27-Mar-1998
Copyright (C) 1998, Thomas G. Lane
Start of Image
Define Huffman Table 0x00
Start Of Frame 0xc3: width=256, height=256, components=1
Component 0: 1hx1v q=0
Start Of Scan: 1 components
Component 0: dc=0 ac=0
Ss=1, Se=0, Ah=0, Al=0
monochrome
lossless (sequential)
Huffman

trying to decompress the image I am getting:

$ ./bin/djpeg16 -outfile dummy.pnm /tmp/GENESIS_SIGNA-JPEG-
CorruptFrag.jpg
Corrupt JPEG data: 121 extraneous bytes before marker 0xd9

So what's again your question? Yes, you can break JPEG files, and
actually any other stream as well. (-;

So long,
Thomas
mathieu
Posted: Thu Apr 17, 2008 11:17 pm
Guest
On Apr 17, 9:16 pm, Thomas Richter <t...@math.tu-berlin.de> wrote:
Quote:
mathieu wrote:
I just found yet another broken JPEG file:

http://gdcm.sourceforge.net/thingies/GENESIS_SIGNA-JPEG-CorruptFrag.jpg

jpeginfo gives me:

$ ./bin/jpegdump16 -verbose /tmp/GENESIS_SIGNA-JPEG-
CorruptFrag.jpg
Independent JPEG Group's JPEGDUMP, version 6b 27-Mar-1998
Copyright (C) 1998, Thomas G. Lane
Start of Image
Define Huffman Table 0x00
Start Of Frame 0xc3: width=256, height=256, components=1
Component 0: 1hx1v q=0
Start Of Scan: 1 components
Component 0: dc=0 ac=0
Ss=1, Se=0, Ah=0, Al=0
monochrome
lossless (sequential)
Huffman

trying to decompress the image I am getting:

$ ./bin/djpeg16 -outfile dummy.pnm /tmp/GENESIS_SIGNA-JPEG-
CorruptFrag.jpg
Corrupt JPEG data: 121 extraneous bytes before marker 0xd9

So what's again your question? Yes, you can break JPEG files, and
actually any other stream as well. (-;

I was hoping that someone would jump in, and say: "of course, your
huffman table is off by one", or "no your predictors are wrong..."

This JPEG comes from a DICOM file, and it looks like IMHO that it
might be possible to still read it properly with a properly patched
JPEG library. That's all.

My 2 cents,
-Mathieu
Thomas Richter
Posted: Fri Apr 18, 2008 3:43 pm
Guest
mathieu wrote:
Quote:
On Apr 17, 9:16 pm, Thomas Richter <t...@math.tu-berlin.de> wrote:
mathieu wrote:
I just found yet another broken JPEG file:
http://gdcm.sourceforge.net/thingies/GENESIS_SIGNA-JPEG-CorruptFrag.jpg
jpeginfo gives me:
$ ./bin/jpegdump16 -verbose /tmp/GENESIS_SIGNA-JPEG-
CorruptFrag.jpg
Independent JPEG Group's JPEGDUMP, version 6b 27-Mar-1998
Copyright (C) 1998, Thomas G. Lane
Start of Image
Define Huffman Table 0x00
Start Of Frame 0xc3: width=256, height=256, components=1
Component 0: 1hx1v q=0
Start Of Scan: 1 components
Component 0: dc=0 ac=0
Ss=1, Se=0, Ah=0, Al=0
monochrome
lossless (sequential)
Huffman
trying to decompress the image I am getting:
$ ./bin/djpeg16 -outfile dummy.pnm /tmp/GENESIS_SIGNA-JPEG-
CorruptFrag.jpg
Corrupt JPEG data: 121 extraneous bytes before marker 0xd9
So what's again your question? Yes, you can break JPEG files, and
actually any other stream as well. (-;

I was hoping that someone would jump in, and say: "of course, your
huffman table is off by one", or "no your predictors are wrong..."

This JPEG comes from a DICOM file, and it looks like IMHO that it
might be possible to still read it properly with a properly patched
JPEG library. That's all.

After looking at it, it likely requires more than a patch. This file
claims to be lossless JPEG (not JPEG-LS, which is again a different
beast). The IJG code cannot handle this JPEG mode, and it's rather
uncommon. There's no DCT in it, it uses an integer transformation.

So long,
Thomas
 
Page 1 of 1       All times are GMT - 5 Hours
The time now is Fri Sep 05, 2008 4:24 pm