[FFmpeg-devel] [PATCH 2/4 v6] avformat: add a Tile Grid stream group type
Andreas Rheinhardt
andreas.rheinhardt at outlook.com
Wed Jan 31 20:08:35 EET 2024
James Almer:
> This will be used to support tiled image formats like HEIF.
>
> Signed-off-by: James Almer <jamrial at gmail.com>
> ---
> libavformat/avformat.c | 5 +++
> libavformat/avformat.h | 100 +++++++++++++++++++++++++++++++++++++++++
> libavformat/dump.c | 29 ++++++++++++
> libavformat/options.c | 32 +++++++++++++
> 4 files changed, 166 insertions(+)
>
> diff --git a/libavformat/avformat.c b/libavformat/avformat.c
> index 8e8c6fbe55..32ef440207 100644
> --- a/libavformat/avformat.c
> +++ b/libavformat/avformat.c
> @@ -100,6 +100,11 @@ void ff_free_stream_group(AVStreamGroup **pstg)
> av_iamf_mix_presentation_free(&stg->params.iamf_mix_presentation);
> break;
> }
> + case AV_STREAM_GROUP_PARAMS_TILE_GRID:
> + av_opt_free(stg->params.tile_grid);
> + av_freep(&stg->params.tile_grid->offsets);
> + av_freep(&stg->params.tile_grid);
> + break;
> default:
> break;
> }
> diff --git a/libavformat/avformat.h b/libavformat/avformat.h
> index 5d0fe82250..6577f13ef1 100644
> --- a/libavformat/avformat.h
> +++ b/libavformat/avformat.h
> @@ -1018,10 +1018,109 @@ typedef struct AVStream {
> int pts_wrap_bits;
> } AVStream;
>
> +/**
> + * AVStreamGroupTileGrid holds information on how to combine several
> + * independent images on a single grid for presentation. None of the tiles may
> + * overlap inside the grid.
> + *
> + * The following is an example of a simple grid with 3 rows and 4 columns:
> + *
> + * +---+---+---+---+
> + * | 0 | 1 | 2 | 3 |
> + * +---+---+---+---+
> + * | 4 | 5 | 6 | 7 |
> + * +---+---+---+---+
> + * | 8 | 9 |10 |11 |
> + * +---+---+---+---+
> + *
> + * Assuming all tiles have a dimension of 512x512, the
> + * @ref AVStreamGroupTileGrid.offsets "offset" of the topleft pixel of
> + * the first @ref AVStreamGroup.streams "stream" in the group is "0,0", the
> + * @ref AVStreamGroupTileGrid.offsets "offset" of the topleft pixel of
> + * the second @ref AVStreamGroup.streams "stream" in the group is "512,0", the
> + * @ref AVStreamGroupTileGrid.offsets "offset" of the topleft pixel of
> + * the fifth @ref AVStreamGroup.streams "stream" in the group is "0,512", the
> + * @ref AVStreamGroupTileGrid.offsets "offset", of the topleft pixel of
> + * the sixth @ref AVStreamGroup.streams "stream" in the group is "512,512",
> + * etc.
> + *
> + * sizeof(AVStreamGroupTileGrid) is not a part of the ABI. No new fields may be
> + * added to this struct without a major version bump.
This is inconsistent. I think you mean that sizeof is part of the ABI.
> + */
> +typedef struct AVStreamGroupTileGrid {
> + const AVClass *av_class;
> +
> + /**
> + * Width of the final image in the grid.
> + *
> + * Must be > 0.
> + */
> + int coded_width;
> + /**
> + * Width of the final image in the grid.
> + *
> + * Must be > 0.
> + */
> + int coded_height;
> +
> + /**
> + * An @ref AVStreamGroup.nb_streams "nb_streams" sized array of offsets in
> + * pixels from the topleft edge of the grid, indicating where each stream
> + * should be placed.
> + * It must be allocated with the av_malloc() family of functions.
> + *
> + * - demuxing: set by libavformat, must not be modified by the caller.
> + * - muxing: set by the caller before avformat_write_header().
> + *
> + * Freed by libavformat in avformat_free_context().
> + */
> + struct {
> + int x;
> + int y;
> + } *offsets;
> +
> + /**
> + * Offset in pixels from the left edge of the grid where the actual image
> + * meant for presentation starts.
> + *
> + * This field must be >= 0 and <= @ref coded_width.
> + */
> + int horizontal_offset;
> + /**
> + * Offset in pixels from the top edge of the grid where the actual image
> + * meant for presentation starts.
> + *
> + * This field must be >= 0 and <= @ref coded_height.
> + */
> + int vertical_offset;
> +
> + /**
> + * Width of the final image for presentation.
> + *
> + * Must be > 0 and <= (@ref coded_width - @ref horizontal_offset).
> + * When it's not equal to (@ref coded_width - @ref horizontal_offset), the
> + * result of (@ref coded_width - width - @ref horizontal_offset) is the
> + * amount amount of pixels to be cropped from the right edge of the
> + * final image before presentation.
> + */
> + int width;
> + /**
> + * Height of the final image for presentation.
> + *
> + * Must be > 0 and <= (@ref coded_height - @ref vertical_offset).
> + * When it's not equal to (@ref coded_height - @ref vertical_offset), the
> + * result of (@ref coded_height - height - @ref vertical_offset) is the
> + * amount amount of pixels to be cropped from the bototm edge of the
> + * final image before presentation.
> + */
> + int height;
> +} AVStreamGroupTileGrid;
> +
> enum AVStreamGroupParamsType {
> AV_STREAM_GROUP_PARAMS_NONE,
> AV_STREAM_GROUP_PARAMS_IAMF_AUDIO_ELEMENT,
> AV_STREAM_GROUP_PARAMS_IAMF_MIX_PRESENTATION,
> + AV_STREAM_GROUP_PARAMS_TILE_GRID,
> };
>
> struct AVIAMFAudioElement;
> @@ -1062,6 +1161,7 @@ typedef struct AVStreamGroup {
> union {
> struct AVIAMFAudioElement *iamf_audio_element;
> struct AVIAMFMixPresentation *iamf_mix_presentation;
> + struct AVStreamGroupTileGrid *tile_grid;
> } params;
>
> /**
> diff --git a/libavformat/dump.c b/libavformat/dump.c
> index aff51b43f6..c9b7369bcd 100644
> --- a/libavformat/dump.c
> +++ b/libavformat/dump.c
> @@ -22,6 +22,7 @@
> #include <stdio.h>
> #include <stdint.h>
>
> +#include "libavutil/avstring.h"
> #include "libavutil/channel_layout.h"
> #include "libavutil/display.h"
> #include "libavutil/iamf.h"
> @@ -721,6 +722,34 @@ static void dump_stream_group(const AVFormatContext *ic, uint8_t *printed,
> }
> break;
> }
> + case AV_STREAM_GROUP_PARAMS_TILE_GRID: {
> + const AVStreamGroupTileGrid *tile_grid = stg->params.tile_grid;
> + AVCodecContext *avctx = avcodec_alloc_context3(NULL);
> + const char *ptr = NULL;
> + av_log(NULL, AV_LOG_INFO, " Tile Grid:");
> + if (avctx && stg->nb_streams && !avcodec_parameters_to_context(avctx, stg->streams[0]->codecpar)) {
> + avctx->width = tile_grid->width;
> + avctx->height = tile_grid->height;
> + avctx->coded_width = tile_grid->coded_width;
> + avctx->coded_height = tile_grid->coded_height;
> + if (ic->dump_separator)
> + av_opt_set(avctx, "dump_separator", ic->dump_separator, 0);
> + buf[0] = 0;
> + avcodec_string(buf, sizeof(buf), avctx, is_output);
> + ptr = av_stristr(buf, " ");
> + }
> + avcodec_free_context(&avctx);
> + if (ptr)
> + av_log(NULL, AV_LOG_INFO, "%s", ptr);
> + av_log(NULL, AV_LOG_INFO, "\n");
> + dump_metadata(NULL, stg->metadata, " ", AV_LOG_INFO);
> + for (int i = 0; i < stg->nb_streams; i++) {
> + const AVStream *st = stg->streams[i];
> + dump_stream_format(ic, st->index, i, index, is_output, AV_LOG_VERBOSE);
> + printed[st->index] = 1;
> + }
> + break;
> + }
> default:
> break;
> }
> diff --git a/libavformat/options.c b/libavformat/options.c
> index 75ec86ce05..ed96ba4afb 100644
> --- a/libavformat/options.c
> +++ b/libavformat/options.c
> @@ -337,6 +337,26 @@ fail:
> return NULL;
> }
>
> +#define FLAGS AV_OPT_FLAG_ENCODING_PARAM | AV_OPT_FLAG_VIDEO_PARAM
> +#define OFFSET(x) offsetof(AVStreamGroupTileGrid, x)
> +static const AVOption tile_grid_options[] = {
> + { "grid_size", "size of the output image", OFFSET(coded_width),
> + AV_OPT_TYPE_IMAGE_SIZE, { .str = NULL }, 0, INT_MAX, FLAGS },
> + { "horizontal_offset", NULL, OFFSET(horizontal_offset), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, FLAGS },
> + { "vertical_offset", NULL, OFFSET(vertical_offset), AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, FLAGS },
> + { "output_size", "size of valid pixels in output image meant for presentation", OFFSET(width),
> + AV_OPT_TYPE_IMAGE_SIZE, { .str = NULL }, 0, INT_MAX, FLAGS },
> + { NULL },
> +};
> +#undef FLAGS
> +#undef OFFSET
> +
> +static const AVClass tile_grid_class = {
> + .class_name = "AVStreamGroupTileGrid",
> + .version = LIBAVUTIL_VERSION_INT,
> + .option = tile_grid_options,
> +};
> +
> static void *stream_group_child_next(void *obj, void *prev)
> {
> AVStreamGroup *stg = obj;
> @@ -346,6 +366,8 @@ static void *stream_group_child_next(void *obj, void *prev)
> return stg->params.iamf_audio_element;
> case AV_STREAM_GROUP_PARAMS_IAMF_MIX_PRESENTATION:
> return stg->params.iamf_mix_presentation;
> + case AV_STREAM_GROUP_PARAMS_TILE_GRID:
> + return stg->params.tile_grid;
> default:
> break;
> }
> @@ -368,6 +390,9 @@ static const AVClass *stream_group_child_iterate(void **opaque)
> case AV_STREAM_GROUP_PARAMS_IAMF_MIX_PRESENTATION:
> ret = av_iamf_mix_presentation_get_class();
> break;
> + case AV_STREAM_GROUP_PARAMS_TILE_GRID:
> + ret = &tile_grid_class;
> + break;
> default:
> break;
> }
> @@ -429,6 +454,13 @@ AVStreamGroup *avformat_stream_group_create(AVFormatContext *s,
> if (!stg->params.iamf_mix_presentation)
> goto fail;
> break;
> + case AV_STREAM_GROUP_PARAMS_TILE_GRID:
> + stg->params.tile_grid = av_mallocz(sizeof(*stg->params.tile_grid));
> + if (!stg->params.tile_grid)
> + goto fail;
> + stg->params.tile_grid->av_class = &tile_grid_class;
> + av_opt_set_defaults(stg->params.tile_grid);
> + break;
> default:
> goto fail;
> }
More information about the ffmpeg-devel
mailing list