[FFmpeg-devel] [PATCH] [MXF] - Add jpeg2000 subdescriptor in MXF file (V2).

[Tomas Härdin]

fre 2023-12-22 klockan 09:58 +0100 skrev Cédric Le Barz:
> 
> Le 21/12/2023 à 12:11, Tomas Härdin a écrit :
> > ons 2023-12-20 klockan 18:04 +0100 skrev Cédric Le Barz:
> > > Le 19/12/2023 à 14:36, Tomas Härdin a écrit :
> > > > > +    for ( comp = 0; comp < component_count; comp++ ) {
> > > > > +        avio_write(pb, &sc->j2k_info.j2k_comp_desc[3*comp] ,
> > > > > 3);
> > > > > +    }
> > > > Looks like this could be simplified to just
> > > > 
> > > >     avio_write(pb, sc->j2k_info.j2k_comp_desc,
> > > > 3*component_count);
> > > > 
> > > > > +    if (j2k_ncomponents != component_count) {
> > > > > +        av_log(s, AV_LOG_ERROR, "Incoherence about
> > > > > components
> > > > > image
> > > > > number.\n");
> > > > > +    }
> > > > I again feel this should be a hard error
> > > > 
> > > > > +    for (comp = 0; comp < j2k_ncomponents; comp++) {
> > > > > +        sc->j2k_info.j2k_comp_desc[comp*j2k_ncomponents] =
> > > > > bytestream2_get_byteu(&g);   // Bitdepth for each component
> > > > > +        sc->j2k_info.j2k_comp_desc[comp*j2k_ncomponents+1] =
> > > > > bytestream2_get_byteu(&g); // Horizontal sampling for each
> > > > > component
> > > > > +        sc->j2k_info.j2k_comp_desc[comp*j2k_ncomponents+2] =
> > > > > bytestream2_get_byteu(&g); // Vertical sampling for each
> > > > > component
> > > > > +    }
> > > > Could be simplified to a single avio_read()
> > > > 
> > > > /Tomas
> > > > _______________________________________________
> > > > ffmpeg-devel mailing list
> > > > ffmpeg-devel at ffmpeg.org
> > > > https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
> > > > 
> > > > To unsubscribe, visit link above, or email
> > > > ffmpeg-devel-request at ffmpeg.org with subject "unsubscribe".
> > > Here is the new version of the patch taken into account your
> > > remarks.
> > Looks OK. Does this also work with RGB(A)? Seems to not be
> > hardcoded
> > for YUV at least. Higher bitdepths would also be nice, I've been
> > working with lossless RGB48 J2K. Doesn't necessarily need to hold
> > up
> > this patch, just curious.
> > 
> > /Tomas
> 
> Any J2K bitdepth can be handled with this patch (up to 31 bits per 
> component).
> 
> Concerning RGBA, The JPEG 2000 sub-descriptor can be referenced
> either 
> by the CDCI picture essence descriptor or the RGBA picture essence 
> descriptor both of which are defined by SMPTE ST 377-1. In this patch
> it 
> is only referenced by the CDCI as the RGBA essence descriptor is not
> yet 
> implemented in FFmpeg.   Nevertheless, I think it should work too
> with 
> CDCI essence descriptor : if you have such data, you can try.

I have one sample at least, muxed by WidgetCo's as-02-wrap. Its
descriptor looks like this (via mxfdump):

RGBAEssenceDescriptor
  InstanceUID = {c3bf8d8c-6fd4-4004-83d7-3950ab56e86f}
  SubDescriptors
    SubDescriptor = {d2013379-ad4b-4d9e-9531-c3a6d8d59175}
    SubDescriptor -> Strong Reference to JPEG2000PictureSubDescriptor
      JPEG2000PictureSubDescriptor
        InstanceUID = {d2013379-ad4b-4d9e-9531-c3a6d8d59175}
        Rsiz = 16384
        Xsiz = 3840
        Ysiz = 2160
        XOsiz = 0
        YOsiz = 0
        XTsiz = 3840
        YTsiz = 2160
        XTOsiz = 0
        YTOsiz = 0
        Csiz = 3
        PictureComponentSizing
          PictureComponentSize = Ssiz="15", XRsiz="1", YRsiz="1"
          PictureComponentSize = Ssiz="15", XRsiz="1", YRsiz="1"
          PictureComponentSize = Ssiz="15", XRsiz="1", YRsiz="1"
        CodingStyleDefault = Scod="0", SGcod="ProgressionOrder="2",
NumberOfLayers="1", MultipleComponentTransformation="1"",
SPcod="DecompositionLevels="5", CodeblockWidth="4",
CodeblockHeight="4", CodeblockStyle="64", Transformation="1"",
PrecinctSize=""
        QuantizationDefault = Sqcd="32", SPqcd="152, 160, 160, 168,
160, 160, 168, 160, 160, 168, 160, 160, 160, 152, 152, 160"
        Unknown ff.e9 [060e2b34.0101.010e.04010603.0e000000] = 52 10 47
10 42 10 00 00 00 00 00 00 00 00 00 00
        Unknown ff.e8 [060e2b34.0101.010e.04010603.0f000000] = 00 02 00
00 00 00 00 01 00 00 00 02 00 02
  LinkedTrackID = 2
  SampleRate = 24/1
  ContainerDuration = 1
  EssenceContainer = MXF-GC JPEG-2000 Picture Mappings
  FrameLayout = 0
  StoredWidth = 3840
  StoredHeight = 2160
  AspectRatio = 3840/2160
  Gamma = [060e2b34.0401.010d.04010101.010a0000]
  PictureEssenceCoding = [060e2b34.0401.010d.04010202.03010801]
  Unknown 32.19 [060e2b34.0101.0109.04010201.01060100] = 06 0e 2b 34 04
01 01 0d 04 01 01 01 03 04 00 00
  Unknown ff.fe [060e2b34.0101.010e.04010501.14000000] = 00 00 00 02
  Unknown ff.fd [060e2b34.0101.010e.04010501.13000000] = 00 00 00 03
  Unknown ff.fc [060e2b34.0101.010e.04010501.15000000] = 00 00 00 04
  Unknown ff.fb [060e2b34.0101.010e.04010501.16000000] = 00 00 00 05
  VideoLineMap
    VideoLineMapEntry = <Unknown>
    VideoLineMapEntry = <Unknown>
  Unknown ff.fa [060e2b34.0101.010e.04200401.01010000] = 7d 00 40 74 3a
98 75 30 1d 4c 0b b8
  Unknown ff.f9 [060e2b34.0101.010e.04200401.01020000] = 3d 13 40 42
  Unknown ff.f8 [060e2b34.0101.010e.04200401.01030000] = 00 98 96 80
  Unknown ff.f7 [060e2b34.0101.010e.04200401.01040000] = 00 00 00 32
  ComponentMaxRef = 65535
  ComponentMinRef = 0
  ScanningDirection = 0
  PixelLayout = <Unknown>

A client has received samples which are multiple MXFs, each one
containing just a single j2k image.

Anyway I will push this patch later today,

/Tomas