[FFmpeg-devel] Adding ICC profile support to MOV decode/encode

[Hello Vectronic]

Hello,

I need to implement support for reading and writing ICC profiles which can be stored in MOV/MP4 sample descriptor colour information.

The relevant extract from the ISO standard is:

class ColourInformationBox extends Box('colr'){
    unsigned int(32) colour_type;
    if (colour_type == 'nclx')  /* on-screen colours */
    {
        unsigned int(16) colour_primaries;
        unsigned int(16) transfer_characteristics;
        unsigned int(16) matrix_coefficients;
        unsigned int(1)  full_range_flag;
        unsigned int(7)  reserved = 0;
    }
    else if (colour_type == 'rICC')
    {
        ICC_profile;    // restricted ICC profile
    }
    else if (colour_type == 'prof')
    {
        ICC_profile;    // unrestricted ICC profile
    }
}

At the moment the code only supports nclc/nclx colour type in:

libavformat/mov.c  => mov_read_colr()
libavformat/moveenc.c =>  mov_write_colr_tag()

Support for ICC profile is implemented on a per frame basis in AVFrameSideDataType.AV_FRAME_DATA_ICC_PROFILE. This is used by the PNG, WEBP and MJPEG codec implementations. The ICC profile in this scenario is treated as an opaque octet buffer. I don't believe this is relevant to my use case as the colour information within the MOV/MP4 sample descriptor relates to the entire stream and is not used on a per-frame basis.

My thinking is to implement support for ICC profile in a similar way that color_range, color_primaries etc. are currently handled:

1. Store the ICC profile as an opaque octet buffer stored in an AVCodecParameters struct value:

uint8_t *icc_profile_data;
int 	icc_profile_size;

2. Add handling of these in libavcodec/utils.c to the following:

codec_parameters_reset()
avcodec_parameters_to_context()
avcodec_parameters_from_context()

2. Add support for reading and writing to the existing MOV/MP4 implementations:

libavformat/mov.c => mov_read_colr()
libavformat/moveenc.c => mov_write_colr_tag()

I'd appreciate any suggestions on better/alternative ways to implement this as I would like to eventually submit as a patch.

Thanks very much,

vectronic