Error decoding JP2 files
Fonzy
We have a device that generates JPEG 2000 files and we have been able to decode the files with other decoders. However, the openjpeg decoder crashes trying to open some of our files. About 1 out of 10 work and the one file that does decode seems to be in black and white rather than color.
I'm new to this group, library and the nomenclature of the JPEG 2000 format so please be gentle with your responses. I can post the code I use to decode the images if you like. Perhaps I have an error in my code but the command line apps are throwing errors as well.
I've attached a sample file that causes issues.
Thanks
Mathieu Malaterre
For some reason, nothing got attached to this email. Could you please
fill in a bug report and attach your sample data. Simply go to:
http://code.google.com/p/openjpeg/issues/list
click 'new issue'
Thanks,
--
Mathieu
Fonzy
Sure no problem. The image not posting is an issue on our side with our firewall. I'll try and fill one out in a bit.
Thanks
On Monday, March 5, 2012 3:46:33 AM UTC-5, mathieu wrote:
Fonzy,
On Monday, March 5, 2012 3:46:33 AM UTC-5, mathieu wrote:
Fonzy,
Mathieu Malaterre
1. openjpeg 1.5 reports:
[ERROR] 00001f2e: expected a marker instead of 0
[ERROR] Failed to decode J2K image
ERROR -> j2k_to_image: failed to decode image!
2. jasper seems to be producing an RGB dataset
3. kakadu reports:
Kakadu Warning:
Trying to save uncommon 3-colour space to TIFF file (JP2 colour space
identifier is 18). The current TIFF writer module will record this as an RGB
space, possibly eroneously.
Output file is grayscale.
> --
> You are subscribed to the mailing-list of the OpenJPEG project
> (www.openjpeg.org)
> To post: email to open...@googlegroups.com
> To unsubscribe: email to openjpeg+u...@googlegroups.com
> For more options: visit http://groups.google.com/group/openjpeg
> For more info about OpenJPEG main partners : visit www.tele.ucl.ac.be (UCL)
> and www.intopix.com
--
Mathieu
mukund yk
Sorry to be crossposting here. I'm not quite used to the concept of a
mailing list. I'm not sure if i can reply to another person's query.
Please let me know if i need to post in a separate thread.
Well, i seem to have the same problem as Fonzy. OpenJPEG reports
something very similar to this:
[ERROR] 00001f2e: expected a marker instead of 0
[ERROR] Failed to decode J2K image
ERROR -> j2k_to_image: failed to decode image!
Thanks!
On Mar 5, 8:43 pm, Mathieu Malaterre <mathieu.malate...@gmail.com>
wrote:
szukw000
FILE(file3307217.jp2)
LENG(7984)
[1]BOX_NAME(ftyp) BOX_LEN(20)
brand(jp2 ) minv(0)
CL[0](jp2 )
[1]BOX_NAME(jp2h) BOX_LEN(45)
[1]BOX_NAME(ihdr) BOX_LEN(22)
read_ihdr
w(720) h(243) nc(3)
signed(0) depth(8)
compress(7) unknown_c(0) ipr(0)
[1]BOX_NAME(colr) BOX_LEN(15)
read_colr
meth(1) prec(0) approx(0) enumcs[18](sYCC)
[1]BOX_NAME(jp2c) BOX_LEN(0)
ENTER read_jp2c
[0]marker(0xff4f)
soc len(0)
[2]marker(0xff51)
siz len(47)
[51]marker(0xff52)
cod len(12)
[65]marker(0xff5c)
qcd len(35)
[102]marker(0xff90)
sot tile_nr(0) Psot(7783) TPsot(0) TNsot(1)
len(10)
[114]marker(0xff93)
sod len(7781)
[7897]marker(0)
MARKER 0 is unknown. STOP.
EXIT read_jp2c
end - s ==> 0
EXIT with end - s ==> 0
winfried
Arnaud Maye
" expected a marker instead of 0 [ERROR] Failed to decode J2K image ERROR ->
j2k_to_image: failed to decode image!"
This is the usual codestream with padding problem I assume, try to get rid
of a few zeroes at the end of initial file. As seen from OpenJPEG the last
two bytes in a code stream is a marker and not zero padding.
One of my patch in the past would take care of this, but removing padding
bytes in the initial file will quickly tell!
Best Regards
Arnaud
Hi Mathieu,
mukund yk
Thanks a lot for the info. The decode works fine after removing the
last two bytes with a marker(0xffd9). Could you please provide more
information as to why this problem occurs? The last 2 bytes being
wrong?
Thanks.
Arnaud Maye
In fact hardware based codecs as ADV212 (Analog devices) tend to produce
padded code stream where it becomes handy to have correct number of bytes
(always falling on a perfect multiple of x). This helps for transferring
data, the idea is to pad the code stream with extra zeroes. I don't have my
JPEG2000 specification handy I am not sure what the standard says about
that. Either the hardware encoder are right or the software frontend should
save files without the padding or jpeg2000 code can be modified in order to
"discard" zero bytes.
Over years we have seen quite a few people having the same issue and I
believe it might be interesting to get oJPEG2000 parser more robust in this
respect. In the other hand, if the spec says last bytes should be the
marker, then oJPEG2000 library is correct and nothing has to change.
Probably the current OPENJPEG2000 manager might indicate his wishes. On my
side I see no problem having the parser improved and this regardless of what
the specification says.
With my best regards,
Arnaud Maye
Hello Arnaud,
Thanks.
For more options: visit http://groups.google.com/group/openjpeg
Fonzy
Sorry I was away from the discussion on this. We are in fact using a Analog Devices ADV212 encoder and therefore we do not control the file that gets generated. I agree with Arnaud, although it may not be 100% to the spec having the library support these padded bytes, is better than the alternative. The alternative is having many people who need a JPEG2000 decoder rediscover this same issue and figure out that they will have to open the file and figure out how to tweak it so it can be used by the library. Keep in mind that in order to tweak the file, the developer will have to understand the JP2 format in order to know whether or not the data should be adjusted which is a duplication of effort.
In other words, OpenJPEG is already decoding the data, why should others have to decode it to determine whether it should be tweaked so the decoder could handle it properly?
In addition, why is it reading the data as grayscale? The image is a color image.
Thanks so much for following up on this issue
> > > For more options: visithttp://groups.google.com/group/openjpeg
> > > For more info about OpenJPEG main partners :
> > > visitwww.tele.ucl.ac.be(UCL) andwww.intopix.com
>
> > --
> > Mathieu
>
> --
> You are subscribed to the mailing-list of the OpenJPEG project
> (www.openjpeg.org) To post: email to open...@googlegroups.com To
> unsubscribe: email to openjpeg+unsubscribe@googlegroups.com
> For more options: visithttp://groups.google.com/group/openjpeg
> For more info about OpenJPEG main partners : visit
> www.tele.ucl.ac.be(UCL) andwww.intopix.com
--
You are subscribed to the mailing-list of the OpenJPEG project
(www.openjpeg.org) To post: email to open...@googlegroups.com To
unsubscribe: email to openjpeg+unsubscribe@googlegroups.com
Arnaud Maye
Hello,
I think you are following my steps but unfortunately most of the patchwork I had done have never been integrated properly. We have been developing an FPGA daughter board, VID475, having two ADV212 codecs.
One ADV for Y compression and one ADV for C compression, then both code streams would be merged resulting in a .jp2 color codestream, I’ve also been able to generate proper MJ2 files at the time. Try to rename your j2k extension as jp2 and open with kakadu, do you have color that way?
Not sure how far you are from what I had been doing but I can share some of the VID475 code ( typically ADV212 configuration as well as merging and MJ2 files creation ).
Best Regards,
Arnaud
> > > For more options: visithttp://groups.google.com/group/openjpeg
> > > For more info about OpenJPEG main partners :
> > > visitwww.tele.ucl.ac.be(UCL) andwww.intopix.com
>
> > --
> > Mathieu
>
> --
> You are subscribed to the mailing-list of the OpenJPEG project
> (www.openjpeg.org) To post: email to open...@googlegroups.com To
> unsubscribe: email to openjpeg+u...@googlegroups.com
> For more options: visithttp://groups.google.com/group/openjpeg
> For more info about OpenJPEG main partners : visit
> www.tele.ucl.ac.be(UCL) andwww.intopix.com
--
You are subscribed to the mailing-list of the OpenJPEG project
(www.openjpeg.org) To post: email to open...@googlegroups.com To
unsubscribe: email to openjpeg+u...@googlegroups.com
For more options: visit http://groups.google.com/group/openjpeg
For more info about OpenJPEG main partners : visit www.tele.ucl.ac.be (UCL)
and www.intopix.com
--
You are subscribed to the mailing-list of the OpenJPEG project (www.openjpeg.org)
To post: email to open...@googlegroups.com
To unsubscribe: email to openjpeg+u...@googlegroups.com
Fonzy
The images work fine with Kakadu and MainConcepts. Jasper doesn't trip over them as well but we had performance issues with Jasper taking too long to decode. If you know of anything we can adjust on the ADV212 chip that would fix the issue would be great. We are not using MJ2 we are sending a series of JP2 files that we display in succession so the end result is basically the same.
Thanks
> > > For more options: visithttp://groups.google.com/group/openjpeg
> > > For more info about OpenJPEG main partners :
> > > visitwww.tele.ucl.ac.be(UCL) andwww.intopix.com
>
> > --
> > Mathieu
>
> --
> You are subscribed to the mailing-list of the OpenJPEG project
> (www.openjpeg.org) To post: email to open...@googlegroups.com To
> unsubscribe: email to openjpeg+unsubscribe@googlegroups.com
> For more options: visithttp://groups.google.com/group/openjpeg
> For more info about OpenJPEG main partners : visit
> www.tele.ucl.ac.be(UCL) andwww.intopix.com
--
You are subscribed to the mailing-list of the OpenJPEG project
(www.openjpeg.org) To post: email to open...@googlegroups.com To
unsubscribe: email to openjpeg+unsubscribe@googlegroups.com
For more options: visit http://groups.google.com/group/openjpeg
For more info about OpenJPEG main partners : visit www.tele.ucl.ac.be (UCL)
and www.intopix.com
--
You are subscribed to the mailing-list of the OpenJPEG project (www.openjpeg.org)
To post: email to open...@googlegroups.com
To unsubscribe: email to openjpeg+unsubscribe@googlegroups.com
Arnaud Maye
Hello Again,
Can you send us your image? I do not mind to look into that, I had any planned to try the new ojpg2k codebase ( I used v1.3 back in 2008 ) oh gosh am getting old…
Thanks,
Arnaud
From: open...@googlegroups.com [mailto:open...@googlegroups.com] On Behalf Of Fonzy
Sent: Thursday, March 15, 2012 2:00 PM
To: open...@googlegroups.com
Subject: Re: [OpenJPEG] Re: Error decoding JP2 files
Arnaud,
The images work fine with Kakadu and MainConcepts. Jasper doesn't trip over them as well but we had performance issues with Jasper taking too long to decode. If you know of anything we can adjust on the ADV212 chip that would fix the issue would be great. We are not using MJ2 we are sending a series of JP2 files that we display in succession so the end result is basically the same.
Thanks
On Thursday, March 15, 2012 8:53:26 AM UTC-4, Arnaud Maye wrote:
Hello,
I think you are following my steps but unfortunately most of the patchwork I had done have never been integrated properly. We have been developing an FPGA daughter board, VID475, having two ADV212 codecs.
One ADV for Y compression and one ADV for C compression, then both code streams would be merged resulting in a .jp2 color codestream, I’ve also been able to generate proper MJ2 files at the time. Try to rename your j2k extension as jp2 and open with kakadu, do you have color that way?
Not sure how far you are from what I had been doing but I can share some of the VID475 code ( typically ADV212 configuration as well as merging and MJ2 files creation ).
Best Regards,
Arnaud
> > > For more options: visithttp://groups.google.com/group/openjpeg
> > > For more info about OpenJPEG main partners :
> > > visitwww.tele.ucl.ac.be(UCL) andwww.intopix.com
>
> > --
> > Mathieu
>
> --
> You are subscribed to the mailing-list of the OpenJPEG project
> (www.openjpeg.org) To post: email to open...@googlegroups.com To
> unsubscribe: email to openjpeg+u...@googlegroups.com
> For more options: visithttp://groups.google.com/group/openjpeg
> For more info about OpenJPEG main partners : visit
> www.tele.ucl.ac.be(UCL) andwww.intopix.com
--
You are subscribed to the mailing-list of the OpenJPEG project
(www.openjpeg.org) To post: email to open...@googlegroups.com To
unsubscribe: email to openjpeg+u...@googlegroups.com
For more options: visit http://groups.google.com/group/openjpeg
For more info about OpenJPEG main partners : visit www.tele.ucl.ac.be (UCL)
and www.intopix.com
--
You are subscribed to the mailing-list of the OpenJPEG project (www.openjpeg.org)
To post: email to open...@googlegroups.com
To unsubscribe: email to openjpeg+u...@googlegroups.com
For more options: visit http://groups.google.com/group/openjpeg
For more info about OpenJPEG main partners : visit www.tele.ucl.ac.be (UCL) and www.intopix.com
--
You are subscribed to the mailing-list of the OpenJPEG project (www.openjpeg.org)
To post: email to open...@googlegroups.com
To unsubscribe: email to openjpeg+u...@googlegroups.com
Fonzy
Here you go. This one fails.
Joe
> > > For more options: visithttp://groups.google.com/group/openjpeg
> > > For more info about OpenJPEG main partners :
> > > visitwww.tele.ucl.ac.be(UCL) andwww.intopix.com
>
> > --
> > Mathieu
>
> --
> You are subscribed to the mailing-list of the OpenJPEG project
> (www.openjpeg.org) To post: email to open...@googlegroups.com To
> unsubscribe: email to openjpeg+unsubscribe@googlegroups.com
> For more options: visithttp://groups.google.com/group/openjpeg
> For more info about OpenJPEG main partners : visit
> www.tele.ucl.ac.be(UCL) andwww.intopix.com
--
You are subscribed to the mailing-list of the OpenJPEG project
(www.openjpeg.org) To post: email to open...@googlegroups.com To
unsubscribe: email to openjpeg+unsubscribe@googlegroups.com
For more options: visit http://groups.google.com/group/openjpeg
For more info about OpenJPEG main partners : visit www.tele.ucl.ac.be (UCL)
and www.intopix.com
--
You are subscribed to the mailing-list of the OpenJPEG project (www.openjpeg.org)
To post: email to open...@googlegroups.com
To unsubscribe: email to openjpeg+unsubscribe@googlegroups.com
For more options: visit http://groups.google.com/group/openjpeg
For more info about OpenJPEG main partners : visit www.tele.ucl.ac.be (UCL) and www.intopix.com
--
You are subscribed to the mailing-list of the OpenJPEG project (www.openjpeg.org)
To post: email to open...@googlegroups.com
To unsubscribe: email to openjpeg+unsubscribe@googlegroups.com
Callum Lerwick
> Sorry I was away from the discussion on this. We are in fact using a
> Analog Devices ADV212 encoder and therefore we do not control the file that
> gets generated. I agree with Arnaud, although it may not be 100% to the
> spec having the library support these padded bytes, is better than the
> alternative. The alternative is having many people who need a JPEG2000
> decoder rediscover this same issue and figure out that they will have to
> open the file and figure out how to tweak it so it can be used by the
> library. Keep in mind that in order to tweak the file, the developer will
> have to understand the JP2 format in order to know whether or not the data
> should be adjusted which is a duplication of effort.
Regardless as to if it meets the letter of the spec or not, if there's
hardware out in the wild outputting padded streams like this, OpenJPEG
really should handle that case gracefully.
Arnaud Maye
I will look into that this week-end if nobody has done that before, it is an easy patch I believe.
--
Sent from my Android phone with K-9 Mail. Please excuse my brevity.
--
You are subscribed to the mailing-list of the OpenJPEG project (www.openjpeg.org)
To post: email to open...@googlegroups.com
To unsubscribe: email to openjpeg+u...@googlegroups.com
szuk...@arcor.de
EOC Required as the last marker in the codestream
=================================================
------- PSOT 0) --------
read_jp2.c:725:
buf_end(0x1d77170) - s(0x1d75317) - 2 ==> PSOT(7767)
box_end(0x1d77170) - s(0x1d75317) ==> PSOT(7767)
sot tile_nr(0) Psot(7767) TPsot(0) TNsot(1)
len(10)
restart at 0x1d7716e
[7881]marker(0)
test_marker:box_end(0x1d77170) - s(0x1d77170) ==> 0
1:MARKER 0 is unknown. STOP.
=================================================
The image does not contain an EOC marker at the end of the
file, but one and only one zero value instead: no padding
zeros.
The resulting images (PNG,TGA,TIF) indeed are colored.
But unsharp: intentionally?
winfried
Arnaud Maye
these files on his side. However I have seen myself padded files (zeroes
after EOC marker) so I am anyway going to patch in this purpose.
-----Original Message-----
From: open...@googlegroups.com [mailto:open...@googlegroups.com] On Behalf
Of szuk...@arcor.de
Sent: Thursday, March 15, 2012 11:41 PM
To: open...@googlegroups.com
winfried
--
Fonzy
The image is unsharp because it is from a standard NTSC source and compressed for a low bit rate. In addition, I took it rather quickly and it was probably a bit out of focus.
The file is generated by the ADV202 chip and thus we have no control over it. However, there are some settings on the chip we will review to see if it can fix the problem. Nonetheless, this padding should be supported by the API.
Thanks
Arnaud Maye
Winfried,
The EOC (FFD9) is there:
00001f10h: 91 7A 61 1E 07 4E D4 7A 0A FF D9 00 00 00 00 00 ; ‘za..NÔz.ÿÙ.....
And the file is padded to the next 8 byte multiple and these are five zeroes at the end of file..
This image is valid or I am missing something here? I would tend to trust more the above line from ultraedit..
Arnaud
From: open...@googlegroups.com [mailto:open...@googlegroups.com] On Behalf Of Fonzy
Sent: Friday, March 16, 2012 2:34 PM
To: open...@googlegroups.com
Subject: Re: [OpenJPEG] Error decoding JP2 files
Winfried,
--
Fonzy
To unsubscribe: email to openjpeg+unsubscribe@googlegroups.com
szukw000
On Mar 16, 3:31 pm, Fonzy <joemac...@gmail.com> wrote:
> I assume Winfried used the OpenJPEG library to generate that report. Could
> it be that there is a bug in the library in which it doesn't find an EOC
> due to some assumption? For example, it may assume the EOC is at the end in
> which case it doesn't see it and it fails. In effect, this isn't a padding
> issue, it is an issue with how the library finds the EOC marker.
library.
The EOC marker is required as the last marker in JP2000 file.
> On Friday, March 16, 2012 10:17:15 AM UTC-4, Arnaud Maye wrote:
>
> > Winfried,
>
> > The EOC (FFD9) is there:
>
> > 00001f10h: 91 7A 61 1E 07 4E D4 7A 0A FF D9 00 00 00 00 00 ;
> > ‘za..NÔz.ÿÙ.....
>
> > And the file is padded to the next 8 byte multiple and these are five
> > zeroes at the end of file..
>
> > This image is valid or I am missing something here? I would tend to trust
> > more the above line from ultraedit..
>
From box_end[-1] to box_end[-7]:
TRACE BOX_END
0, 0, 0, 0, 0,d9,ff
From buf_end[-1] to buf_end[-7]:
TRACE BUF_END
0, 0, 0, 0, 0,d9,ff
This means:
the jp2c box ends on the last byte of the file
the EOC (0xffd9) marker is WITHIN the jp2c box
the jp2c box is padded with 5 zero-bytes AFTER the EOC (0xffd9)
marker
If padding should be allowed then
the jp2c box must end BEFORE the EOC (0xffd9) marker
the padding bytes must follow the EOC (0xffd9) marker
winfried
szukw000
>
> the jp2c box ends on the last byte of the file
>
> the EOC (0xffd9) marker is WITHIN the jp2c box
>
> the jp2c box is padded with 5 zero-bytes AFTER the EOC (0xffd9)
> marker
>
> If padding should be allowed then
>
> the jp2c box must end BEFORE the EOC (0xffd9) marker
in the jp2c box.
So the patch I had sent is perhaps the cheapest solution to get
an image.
winfried
Fonzy
szuk...@arcor.de
No ERROR, no WARNING: quiet.
Two additional tests because of some padding bytes: expensive.
winfried
Arnaud Maye
--
Sent from my Android phone with K-9 Mail. Please excuse my brevity.
--
You are subscribed to the mailing-list of the OpenJPEG project (www.openjpeg.org)
To post: email to open...@googlegroups.com
To unsubscribe: email to openjpeg+u...@googlegroups.com