[FFmpeg-devel] [PATCH v2] avcodec/mjpegenc: support writing ICC profiles
Andreas Rheinhardt
andreas.rheinhardt at outlook.com
Fri Mar 18 17:37:39 EET 2022
Niklas Haas:
> From: Niklas Haas <git at haasn.dev>
>
> This is mostly straightforward. The major complication is that, as a
> result of the 16-bit chunk size limitation, ICC profiles may need to be
> split up into multiple chunks.
>
> We also need to make sure to allocate enough extra space in the packet
> to fit the ICC profile, so modify both mpegvideo_enc.c and ljpegenc.c to
> take into account this extra overhead.
>
> Also add a FATE transcode test to ensure that the ICC profile gets
> written (and read) correctly. Note that this ICC profile is smaller than
> 64 kB, so this doesn't test the APP2 chunk re-arranging code at all.
>
> Signed-off-by: Niklas Haas <git at haasn.dev>
> ---
> Changes in v2:
> - Merge FATE test into this commit
> - Fix possible segfault (when no AVFrame is available)
When could this ever happen? IMO it can't.
> ---
> libavcodec/ljpegenc.c | 6 ++++-
> libavcodec/mjpegenc.c | 3 ++-
> libavcodec/mjpegenc_common.c | 43 +++++++++++++++++++++++++++++++++---
> libavcodec/mjpegenc_common.h | 2 +-
> libavcodec/mpegvideo_enc.c | 3 +++
> tests/fate/image.mak | 3 +++
> tests/ref/fate/jpg-icc | 42 +++++++++++++++++++++++++++++++++++
> 7 files changed, 96 insertions(+), 6 deletions(-)
> create mode 100644 tests/ref/fate/jpg-icc
>
> diff --git a/libavcodec/ljpegenc.c b/libavcodec/ljpegenc.c
> index e15f448f90..c54450c338 100644
> --- a/libavcodec/ljpegenc.c
> +++ b/libavcodec/ljpegenc.c
> @@ -216,6 +216,7 @@ static int ljpeg_encode_frame(AVCodecContext *avctx, AVPacket *pkt,
> {
> LJpegEncContext *s = avctx->priv_data;
> PutBitContext pb;
> + const AVFrameSideData *sd;
> const int width = avctx->width;
> const int height = avctx->height;
> const int mb_width = (width + s->hsample[0] - 1) / s->hsample[0];
> @@ -233,12 +234,15 @@ static int ljpeg_encode_frame(AVCodecContext *avctx, AVPacket *pkt,
> * s->hsample[0] * s->vsample[0];
> }
>
> + if ((sd = av_frame_get_side_data(pict, AV_FRAME_DATA_ICC_PROFILE)))
> + max_pkt_size += sd->size;
This does not account for the chunk overhead.
> +
> if ((ret = ff_alloc_packet(avctx, pkt, max_pkt_size)) < 0)
> return ret;
>
> init_put_bits(&pb, pkt->data, pkt->size);
>
> - ff_mjpeg_encode_picture_header(avctx, &pb, NULL, &s->scantable,
> + ff_mjpeg_encode_picture_header(avctx, &pb, pict, NULL, &s->scantable,
> s->pred, s->matrix, s->matrix);
>
> header_bits = put_bits_count(&pb);
> diff --git a/libavcodec/mjpegenc.c b/libavcodec/mjpegenc.c
> index 08671b0df7..d7c0c763a1 100644
> --- a/libavcodec/mjpegenc.c
> +++ b/libavcodec/mjpegenc.c
> @@ -77,7 +77,8 @@ static av_cold void init_uni_ac_vlc(const uint8_t huff_size_ac[256],
>
> static void mjpeg_encode_picture_header(MpegEncContext *s)
> {
> - ff_mjpeg_encode_picture_header(s->avctx, &s->pb, s->mjpeg_ctx,
> + const AVFrame *frame = s->picture->f;
> + ff_mjpeg_encode_picture_header(s->avctx, &s->pb, frame, s->mjpeg_ctx,
> &s->intra_scantable, 0,
> s->intra_matrix, s->chroma_intra_matrix);
>
> diff --git a/libavcodec/mjpegenc_common.c b/libavcodec/mjpegenc_common.c
> index 995e2b7670..5594a8d239 100644
> --- a/libavcodec/mjpegenc_common.c
> +++ b/libavcodec/mjpegenc_common.c
> @@ -130,8 +130,10 @@ static void jpeg_table_header(AVCodecContext *avctx, PutBitContext *p,
> AV_WB16(ptr, size);
> }
>
> -static void jpeg_put_comments(AVCodecContext *avctx, PutBitContext *p)
> +static void jpeg_put_comments(AVCodecContext *avctx, PutBitContext *p,
> + const AVFrame *frame)
> {
> + const AVFrameSideData *sd = NULL;
> int size;
> uint8_t *ptr;
>
> @@ -161,6 +163,41 @@ static void jpeg_put_comments(AVCodecContext *avctx, PutBitContext *p)
> put_bits(p, 8, 0); /* thumbnail height */
> }
>
> + /* ICC profile */
> + if (frame)
> + sd = av_frame_get_side_data(frame, AV_FRAME_DATA_ICC_PROFILE);
> + if (sd && sd->size) {
> + static const uint8_t hdr_size = strlen("ICC_PROFILE")+5;
This is illegal as initializers for static objects need to be constant
expressions or string literals. Which this thing isn't. You are
basically presuming that strlen be evaluated at compile-time.
> + static const uint16_t max_chunk_size = UINT16_MAX - hdr_size;
> + const uint8_t *data = sd->data;
> + size_t remaining = sd->size;
> + size_t nb_chunks = (remaining + max_chunk_size - 1) / max_chunk_size;
There is a potential overflow here. You could avoid it by checking
whether remaining is > 255 * max_chunk_size.
> + if (nb_chunks > 255) {
> + av_log(avctx, AV_LOG_WARNING,
> + "Cannot store %zu byte ICC profile: too large for JPEG\n",
We use SIZE_SPECIFIER for compatibility with old versions of MSVC. Note
that I do not know whether this is still needed, as all currently
supported MSVC versions might already support #zu.
> + sd->size);
> + nb_chunks = remaining = 0;
> + }
> +
> + for (int i = 0; i < nb_chunks; i++) {
> + size = FFMIN(remaining, max_chunk_size);
> + av_assert1(size > 0);
> + put_marker(p, APP2);
> + put_bits(p, 16, size + hdr_size);
> + ff_put_string(p, "ICC_PROFILE", 1);
> + put_bits(p, 8, i+1);
> + put_bits(p, 8, nb_chunks);
> + flush_put_bits(p);
> + ptr = put_bits_ptr(p);
> + skip_put_bytes(p, size);
> + memcpy(ptr, data, size);
> + remaining -= size;
Everything that you are doing here is byte-aligned, so I think it would
be cleaner (and less code) to write everything directly and
skip_put_bytes once afterwards.
> + data += size;
> + }
> +
> + av_assert1(!remaining);
> + }
> +
> /* comment */
> if (!(avctx->flags & AV_CODEC_FLAG_BITEXACT)) {
> put_marker(p, COM);
> @@ -213,7 +250,7 @@ void ff_mjpeg_init_hvsample(AVCodecContext *avctx, int hsample[4], int vsample[4
> }
>
> void ff_mjpeg_encode_picture_header(AVCodecContext *avctx, PutBitContext *pb,
> - MJpegContext *m,
> + const AVFrame *frame, struct MJpegContext *m,
> ScanTable *intra_scantable, int pred,
> uint16_t luma_intra_matrix[64],
> uint16_t chroma_intra_matrix[64])
> @@ -233,7 +270,7 @@ void ff_mjpeg_encode_picture_header(AVCodecContext *avctx, PutBitContext *pb,
> if (avctx->codec_id == AV_CODEC_ID_AMV)
> return;
>
> - jpeg_put_comments(avctx, pb);
> + jpeg_put_comments(avctx, pb, frame);
>
> jpeg_table_header(avctx, pb, m, intra_scantable,
> luma_intra_matrix, chroma_intra_matrix, hsample);
> diff --git a/libavcodec/mjpegenc_common.h b/libavcodec/mjpegenc_common.h
> index ac753bf153..3c0a7addac 100644
> --- a/libavcodec/mjpegenc_common.h
> +++ b/libavcodec/mjpegenc_common.h
> @@ -30,7 +30,7 @@
> struct MJpegContext;
>
> void ff_mjpeg_encode_picture_header(AVCodecContext *avctx, PutBitContext *pb,
> - struct MJpegContext *m,
> + const AVFrame *frame, struct MJpegContext *m,
> ScanTable *intra_scantable, int pred,
> uint16_t luma_intra_matrix[64],
> uint16_t chroma_intra_matrix[64]);
> diff --git a/libavcodec/mpegvideo_enc.c b/libavcodec/mpegvideo_enc.c
> index c69114ea15..932dfa34f7 100644
> --- a/libavcodec/mpegvideo_enc.c
> +++ b/libavcodec/mpegvideo_enc.c
> @@ -1686,10 +1686,13 @@ int ff_mpv_encode_picture(AVCodecContext *avctx, AVPacket *pkt,
>
> /* output? */
> if (s->new_picture.f->data[0]) {
> + const AVFrameSideData *sd;
> int growing_buffer = context_count == 1 && !pkt->data && !s->data_partitioning;
> int pkt_size = growing_buffer ? FFMAX(s->mb_width*s->mb_height*64+10000, avctx->internal->byte_buffer_size) - AV_INPUT_BUFFER_PADDING_SIZE
> :
> s->mb_width*s->mb_height*(MAX_MB_BYTES+100)+10000;
> + if ((sd = av_frame_get_side_data(s->new_picture.f, AV_FRAME_DATA_ICC_PROFILE)))
> + pkt_size += sd->size;
This does not account for the chunk overhead and it is also not targeted
to MJPEG.
> if ((ret = ff_alloc_packet(avctx, pkt, pkt_size)) < 0)
> return ret;
> if (s->mb_info) {
> diff --git a/tests/fate/image.mak b/tests/fate/image.mak
> index 573d398915..8b72890fbe 100644
> --- a/tests/fate/image.mak
> +++ b/tests/fate/image.mak
> @@ -337,6 +337,9 @@ fate-jpg-12bpp: CMD = framecrc -idct simple -i $(TARGET_SAMPLES)/jpg/12bpp.jpg -
> FATE_JPG += fate-jpg-jfif
> fate-jpg-jfif: CMD = framecrc -idct simple -i $(TARGET_SAMPLES)/jpg/20242.jpg
>
> +FATE_JPG += fate-jpg-icc
> +fate-jpg-icc: CMD = transcode png_pipe $(TARGET_SAMPLES)/png1/lena-int_rgb24.png mjpeg "-vf scale" "" "" "-show_frames"
This needs an ffprobe dependency.
> +
> FATE_JPG-$(call DEMDEC, IMAGE2, MJPEG) += $(FATE_JPG)
> FATE_IMAGE += $(FATE_JPG-yes)
> fate-jpg: $(FATE_JPG-yes)
> diff --git a/tests/ref/fate/jpg-icc b/tests/ref/fate/jpg-icc
> new file mode 100644
> index 0000000000..220146555e
> --- /dev/null
> +++ b/tests/ref/fate/jpg-icc
> @@ -0,0 +1,42 @@
> +0a323df5cdfb9574e329b9831be054a6 *tests/data/fate/jpg-icc.mjpeg
> +11010 tests/data/fate/jpg-icc.mjpeg
> +#tb 0: 1/25
> +#media_type 0: video
> +#codec_id 0: rawvideo
> +#dimensions 0: 128x128
> +#sar 0: 1/1
> +0, 0, 0, 1, 49152, 0xaac06b42
> +[FRAME]
> +media_type=video
> +stream_index=0
> +key_frame=1
> +pts=0
> +pts_time=0.000000
> +pkt_dts=0
> +pkt_dts_time=0.000000
> +best_effort_timestamp=0
> +best_effort_timestamp_time=0.000000
> +pkt_duration=1
> +pkt_duration_time=0.040000
> +pkt_pos=0
> +pkt_size=11010
> +width=128
> +height=128
> +pix_fmt=yuvj444p
> +sample_aspect_ratio=1:1
> +pict_type=I
> +coded_picture_number=0
> +display_picture_number=0
> +interlaced_frame=0
> +top_field_first=0
> +repeat_pict=0
> +color_range=pc
> +color_space=bt470bg
> +color_primaries=unknown
> +color_transfer=unknown
> +chroma_location=center
> +[SIDE_DATA]
> +side_data_type=ICC profile
> +size=3144
> +[/SIDE_DATA]
> +[/FRAME]
More information about the ffmpeg-devel
mailing list