[ERROR] COD marker already read. No more than one COD marker per tile.

[beva]

Hi

I do experience this error from version 2.3.0 of openjpeg when I decompress. The images in focus are originally created from a bulk scanner. I am talking about millions of images. Previous version 2.1.0 of openjepeg do not have any problems with those images. I have included reports from both opj_dump and jpylyzer, both telling that the images are fine. The behaviour is:

beva@vax04:~/Bilder$ opj_decompress -i db11217051900674.jp2 -o test.png

[INFO] Start to read j2k main header (111).
[INFO] Main header has been correctly decoded.
[INFO] No decoded area parameters, set the decoded area to the whole image
[ERROR] COD marker already read. No more than one COD marker per tile.
[ERROR] Fail to read the current marker segment (0xff52)
[ERROR] Failed to decode the codestream in the JP2 file
ERROR -> opj_decompress: failed to decode image!

I then downloaded the source (version 2.3.0) and found the part in the code which is responsible for this behaviour. In src/lib/openjp2/j2k.c at about line 2660 you will find this code snippet:

    /* Only one COD per tile */

    if (l_tcp->cod) {
        opj_event_msg(p_manager, EVT_ERROR,
                      "COD marker already read. No more than one COD marker per tile.\n");
        return OPJ_FALSE;
    }
    l_tcp->cod = 1;

Commenting out this snippet, letting only "l_tcp->cod = 1;" be active, or change "if (l_tcp->cod)" to "if (l_tcp->cod > 1)" and then compile the code, it will decompress my files without any problems.

I am not experienced in C or the JPEG2000 standard, however it seems that the test for multiple COD's may be broken somehow. Could it be that "l_tcp->cod" is not nilled between tiles?

If the test turns out to be logically correct, and thus my files in fact violates the JPEG2000, could it just be better to spawn a warning during decompression, instead of a show stopper action. Further on, in this case, should opj_dump gives a message of violation and not telling that the file is fine.

With very best regards

Bent, Oslo, Norway

Anlyzes_of_jpeg2000_image.txt

[rouault]

Could you provide a link to such an image for further investigations ?

[beva]

I have put an example at dropbox, and even a decompressed png version.

https://www.dropbox.com/sh/kybf40i7v2cfoan/AAC0g1mEIz7w8YPmm-_Nnw5Ja?dl=0

Bent

[rouault]

@mayeut Should have CC'ed you since you authored daed8cc9195555e101ab708a501af2dfe6d5e001which I've basically reverted per 98cfdd3