#785 chromasample for DVCPRO HD is 8 bit, not 10 bit

Incorrect_result
closed-fixed
DV (1)
5
2013-11-18
2013-08-28
cyang
No

To fix it, remove following line 483 of File_DvDif.cpp
Fill(Stream_Video, 0, Video_BitDepth, 10, 10, true); //MXF files say that DVCPRO HD are 10 b

Discussion

    • status: open --> pending
    • assigned_to: Jerome Martinez
     
  • All MXF headers I checked say "Active bits per sample" is "10".
    SMPTE ST 370 says:
    Table 25 – Construction of input video
    1080/60i system 1080/50i system 720/60p system 720/50p system
    Quantization: Each sample is linearly quantized to 10 bits for Y, CR and CB.
    But 4.1.1 says "A sample conversion from 10-bit input video to 8 bits or more is provided by the resampling process (the first processing block of figure 1)."

    I am a bit lost with "or more" and the MXF header saying 10-bit.

    Is DVCPRO HD 8-bit only? or DVCPRO HD 8 or 10-bit, depending of something?
    I try to be able to fill the bit depth field for raw DVCPRO HD, and in the case it is always 8-bit, MediaInfo will display a warning about incoherency between raw stream and MXF header i nthe future, so I need to be sure of my change.

     
  • After deep investigation, it appears that DVCPro does make it possible for 10-bit and 8-bit.

    Saying "10-bit" everywhere (current code) is wrong, saying "DVCPRO HD is 8 bit, not 10 bit" is wrong too.
    i have both in my sample files.
    I'll try to find a quick way to detect bit depth (spec is not obvious), else I remove the piece of information until someone else find the right method.

     
    • status: pending --> open-accepted
     
  • I removed the piece of information from the DV parser.
    I use information from MXF (I have files with 8 and 10 bits) when applicable.

     
    • status: open-accepted --> closed-fixed