[nvJPEG2000]Looks like the HT Decoding not support 12/16bit?

[baronren]

I have made a 16bit HT decoding test, but the result is very strange, the return status is normal, but the file opening is a strange gray image.Here is the result: $ ./nvjpeg2000_decode_sample -i ../images/OmegaCentari_16bit_jpeg2000part15_rate4bpp.j2c -o . Decoding images in directory: ../images/OmegaCentari_16bit_jpeg2000part15_rate4bpp.j2c, total 1, batchsize 1 Total decoding time: 0.00992461 Avg decoding time per image: 0.00992461 Avg images per sec: 100.76 Avg decoding time per batch: 0.00992461 The source image file is,decoded file is

I tried 8bit HT image is OK, looks like that nvJPEG2000 decoder not support above 8bit file. OmegaCentari_16bit_jpeg2000part15_rate4bpp.bmp.zip

[zohebk-nv]

Hi, thank you for reporting this issue. It looks like OmegaCentari_16bit_jpeg2000part15_rate4bpp.j2c has significance propagation and magnitude refinement passes which is not supported by nvJPEG2000 currently. As a work around, would it be possible for you to generate this bitstream with significance propagation and magnitude refinement passes disabled and try it with nvJPEG2000

[michaeldsmith]

The attached zip files contain 8/10/12/16 bit HTJ2K codestreams without any significant propagation and magnitude refinement passes.

OmegaCentari_16bit_HT_Qfactor90.j2c.zip OmegaCentari_12bit_HT_Qfactor90.j2c.zip OmegaCentari_10bit_HT_Qfactor90.j2c.zip OmegaCentari_8bit_HT_Qfactor90.j2c.zip

[zohebk-nv]

Thanks for sharing the HTJ2K files. These codestreams have the QCC marker which is not supported currently by nvJPEG2000. We will look into adding support for the QCC marker.

[trybicki-nv]

Lastest version of nvJPEG2000 (v0.7.5) supports QCC markers.