diff mbox series

[FFmpeg-devel,1/2] avcodec/libjxldec: add animated decode support

Message ID 20230303203146.377726-2-leo.izen@gmail.com
State New
Headers show
Series JPEG XL Animation Support | expand

Checks

Context Check Description
yinshiyou/make_loongarch64 success Make finished
yinshiyou/make_fate_loongarch64 success Make fate finished
andriy/make_x86 success Make finished
andriy/make_fate_x86 success Make fate finished

Commit Message

Leo Izen March 3, 2023, 8:31 p.m. UTC
Migrate the libjxl decoder wrapper from the decode_frame method to the
receive_frame method, which allows sending more than one frame from a
single packet. This allows the libjxl decoder to decode JPEG XL files
that are animated, and emit every frame of the animation. Now, clients
that feed the libjxl decoder with an animated JPEG XL file will be able
to receieve the full animation.

Signed-off-by: Leo Izen <leo.izen@gmail.com>
---
 libavcodec/libjxldec.c | 103 ++++++++++++++++++++++++++++++-----------
 libavcodec/version.h   |   2 +-
 2 files changed, 78 insertions(+), 27 deletions(-)

Comments

Anton Khirnov March 17, 2023, 6:59 p.m. UTC | #1
Quoting Leo Izen (2023-03-03 21:31:45)
> -static int libjxl_decode_frame(AVCodecContext *avctx, AVFrame *frame, int *got_frame, AVPacket *avpkt)
> +static int libjxl_receive_frame(AVCodecContext *avctx, AVFrame *frame)
>  {
>      LibJxlDecodeContext *ctx = avctx->priv_data;
> -    const uint8_t *buf = avpkt->data;
> -    size_t remaining = avpkt->size;
>      JxlDecoderStatus jret;
>      int ret;
> -    *got_frame = 0;
> +    AVPacket *pkt = ctx->avpkt;
> +
> +    if (!pkt->size) {

Shouldn't this live inside the loop?
It won't happen currently, but in principle ff_decode_get_packet() could
return more than one packet.

> +        av_packet_unref(pkt);
> +        ret = ff_decode_get_packet(avctx, pkt);
> +        if (ret < 0 && ret != AVERROR_EOF)
> +            return ret;
> +        ctx->remaining = pkt->size;
> +        if (!pkt->size) {
> +            /* empty packet means eof */
> +            if (ret >= 0) {
> +                av_packet_unref(pkt);
> +                return AVERROR(EAGAIN);

pkt->size == 0 && ret >= 0
should not happen

> +            } else {
> +                return AVERROR_EOF;
> +            }
> +        }
> +    }
>  
>      while (1) {
>  
> -        jret = JxlDecoderSetInput(ctx->decoder, buf, remaining);
> +        jret = JxlDecoderSetInput(ctx->decoder, pkt->data + (pkt->size - ctx->remaining), ctx->remaining);

Wouldn't it be simpler to offset pkt->data and decremented pkt->size?
Then you shouldn't need ctx->remaining at all.

> @@ -419,25 +471,23 @@ static int libjxl_decode_frame(AVCodecContext *avctx, AVFrame *frame, int *got_f
>                  /* ownership is transfered, and it is not ref-ed */
>                  ctx->iccp = NULL;
>              }
> -            *got_frame = 1;
> -            return avpkt->size - remaining;
> +            frame->pts = av_rescale_q(ctx->pts, ctx->timebase, avctx->pkt_timebase);

I think pkt_timebase does not have to be set.
Also, you should set frame->duration.
diff mbox series

Patch

diff --git a/libavcodec/libjxldec.c b/libavcodec/libjxldec.c
index 045a1535f9..394fd8698a 100644
--- a/libavcodec/libjxldec.c
+++ b/libavcodec/libjxldec.c
@@ -52,13 +52,20 @@  typedef struct LibJxlDecodeContext {
 #endif
     JxlDecoderStatus events;
     AVBufferRef *iccp;
+    AVPacket *avpkt;
+    size_t remaining;
+    int64_t pts;
+    int64_t frame_duration;
+    int prev_is_last;
+    AVRational timebase;
 } LibJxlDecodeContext;
 
 static int libjxl_init_jxl_decoder(AVCodecContext *avctx)
 {
     LibJxlDecodeContext *ctx = avctx->priv_data;
 
-    ctx->events = JXL_DEC_BASIC_INFO | JXL_DEC_FULL_IMAGE | JXL_DEC_COLOR_ENCODING;
+    ctx->events = JXL_DEC_BASIC_INFO | JXL_DEC_FULL_IMAGE
+        | JXL_DEC_COLOR_ENCODING | JXL_DEC_FRAME;
     if (JxlDecoderSubscribeEvents(ctx->decoder, ctx->events) != JXL_DEC_SUCCESS) {
         av_log(avctx, AV_LOG_ERROR, "Error subscribing to JXL events\n");
         return AVERROR_EXTERNAL;
@@ -71,6 +78,8 @@  static int libjxl_init_jxl_decoder(AVCodecContext *avctx)
 
     memset(&ctx->basic_info, 0, sizeof(JxlBasicInfo));
     memset(&ctx->jxl_pixfmt, 0, sizeof(JxlPixelFormat));
+    ctx->prev_is_last = 1;
+    ctx->frame_duration = 1;
 
     return 0;
 }
@@ -93,6 +102,11 @@  static av_cold int libjxl_decode_init(AVCodecContext *avctx)
         return AVERROR_EXTERNAL;
     }
 
+    ctx->avpkt = av_packet_alloc();
+    if (!ctx->avpkt)
+        return AVERROR(ENOMEM);
+    ctx->pts = 0;
+
     return libjxl_init_jxl_decoder(avctx);
 }
 
@@ -328,18 +342,33 @@  static int libjxl_color_encoding_event(AVCodecContext *avctx, AVFrame *frame)
     return 0;
 }
 
-static int libjxl_decode_frame(AVCodecContext *avctx, AVFrame *frame, int *got_frame, AVPacket *avpkt)
+static int libjxl_receive_frame(AVCodecContext *avctx, AVFrame *frame)
 {
     LibJxlDecodeContext *ctx = avctx->priv_data;
-    const uint8_t *buf = avpkt->data;
-    size_t remaining = avpkt->size;
     JxlDecoderStatus jret;
     int ret;
-    *got_frame = 0;
+    AVPacket *pkt = ctx->avpkt;
+
+    if (!pkt->size) {
+        av_packet_unref(pkt);
+        ret = ff_decode_get_packet(avctx, pkt);
+        if (ret < 0 && ret != AVERROR_EOF)
+            return ret;
+        ctx->remaining = pkt->size;
+        if (!pkt->size) {
+            /* empty packet means eof */
+            if (ret >= 0) {
+                av_packet_unref(pkt);
+                return AVERROR(EAGAIN);
+            } else {
+                return AVERROR_EOF;
+            }
+        }
+    }
 
     while (1) {
 
-        jret = JxlDecoderSetInput(ctx->decoder, buf, remaining);
+        jret = JxlDecoderSetInput(ctx->decoder, pkt->data + (pkt->size - ctx->remaining), ctx->remaining);
 
         if (jret == JXL_DEC_ERROR) {
             /* this should never happen here unless there's a bug in libjxl */
@@ -353,19 +382,18 @@  static int libjxl_decode_frame(AVCodecContext *avctx, AVFrame *frame, int *got_f
          * of bytes remaining to be read, rather than
          * the number of bytes that it did read
          */
-        remaining = JxlDecoderReleaseInput(ctx->decoder);
-        buf = avpkt->data + avpkt->size - remaining;
+        ctx->remaining = JxlDecoderReleaseInput(ctx->decoder);
 
         switch(jret) {
         case JXL_DEC_ERROR:
             av_log(avctx, AV_LOG_ERROR, "Unknown libjxl decode error\n");
             return AVERROR_INVALIDDATA;
         case JXL_DEC_NEED_MORE_INPUT:
-            if (remaining == 0) {
+            av_log(avctx, AV_LOG_DEBUG, "NEED_MORE_INPUT event emitted\n");
+            if (ctx->remaining == 0) {
                 av_log(avctx, AV_LOG_ERROR, "Unexpected end of JXL codestream\n");
                 return AVERROR_INVALIDDATA;
             }
-            av_log(avctx, AV_LOG_DEBUG, "NEED_MORE_INPUT event emitted\n");
             continue;
         case JXL_DEC_BASIC_INFO:
             av_log(avctx, AV_LOG_DEBUG, "BASIC_INFO event emitted\n");
@@ -384,6 +412,13 @@  static int libjxl_decode_frame(AVCodecContext *avctx, AVFrame *frame, int *got_f
             }
             if ((ret = ff_set_dimensions(avctx, ctx->basic_info.xsize, ctx->basic_info.ysize)) < 0)
                 return ret;
+            if (ctx->basic_info.have_animation) {
+                ctx->timebase = av_make_q(
+                    ctx->basic_info.animation.tps_denominator,
+                    ctx->basic_info.animation.tps_numerator);
+            } else {
+                ctx->timebase = avctx->pkt_timebase;
+            }
             continue;
         case JXL_DEC_COLOR_ENCODING:
             av_log(avctx, AV_LOG_DEBUG, "COLOR_ENCODING event emitted\n");
@@ -407,11 +442,28 @@  static int libjxl_decode_frame(AVCodecContext *avctx, AVFrame *frame, int *got_f
             }
 #endif
             continue;
+        case JXL_DEC_FRAME:
+            av_log(avctx, AV_LOG_DEBUG, "FRAME event emitted\n");
+            if (!ctx->basic_info.have_animation || ctx->prev_is_last) {
+                frame->pict_type = AV_PICTURE_TYPE_I;
+                frame->key_frame = 1;
+            }
+            if (ctx->basic_info.have_animation) {
+                JxlFrameHeader header;
+                if (JxlDecoderGetFrameHeader(ctx->decoder, &header) != JXL_DEC_SUCCESS) {
+                    av_log(avctx, AV_LOG_ERROR, "Bad libjxl dec frame event\n");
+                    return AVERROR_EXTERNAL;
+                }
+                ctx->prev_is_last = header.is_last;
+                ctx->frame_duration = header.duration;
+            } else {
+                ctx->prev_is_last = 1;
+                ctx->frame_duration = 1;
+            }
+            continue;
         case JXL_DEC_FULL_IMAGE:
             /* full image is one frame, even if animated */
             av_log(avctx, AV_LOG_DEBUG, "FULL_IMAGE event emitted\n");
-            frame->pict_type = AV_PICTURE_TYPE_I;
-            frame->key_frame = 1;
             if (ctx->iccp) {
                 AVFrameSideData *sd = av_frame_new_side_data_from_buf(frame, AV_FRAME_DATA_ICC_PROFILE, ctx->iccp);
                 if (!sd)
@@ -419,25 +471,23 @@  static int libjxl_decode_frame(AVCodecContext *avctx, AVFrame *frame, int *got_f
                 /* ownership is transfered, and it is not ref-ed */
                 ctx->iccp = NULL;
             }
-            *got_frame = 1;
-            return avpkt->size - remaining;
+            frame->pts = av_rescale_q(ctx->pts, ctx->timebase, avctx->pkt_timebase);
+            ctx->pts += ctx->frame_duration;
+            return 0;
         case JXL_DEC_SUCCESS:
             av_log(avctx, AV_LOG_DEBUG, "SUCCESS event emitted\n");
             /*
-             * The SUCCESS event isn't fired until after JXL_DEC_FULL_IMAGE. If this
-             * stream only contains one JXL image then JXL_DEC_SUCCESS will never fire.
-             * If the image2 sequence being decoded contains several JXL files, then
-             * libjxl will fire this event after the next AVPacket has been passed,
-             * which means the current packet is actually the next image in the sequence.
-             * This is why we reset the decoder and populate the packet data now, since
-             * this is the next packet and it has not been decoded yet. The decoder does
-             * have to be reset to allow us to use it for the next image, or libjxl
-             * will become very confused if the header information is not identical.
+             * this event will be fired when the zero-length EOF
+             * packet is sent to the decoder by the client,
+             * but it will also be fired when the next image of
+             * an image2pipe sequence is loaded up
              */
             JxlDecoderReset(ctx->decoder);
             libjxl_init_jxl_decoder(avctx);
-            buf = avpkt->data;
-            remaining = avpkt->size;
+            if (!ctx->remaining) {
+                av_packet_unref(pkt);
+                return AVERROR_EOF;
+            }
             continue;
         default:
              av_log(avctx, AV_LOG_ERROR, "Bad libjxl event: %d\n", jret);
@@ -457,6 +507,7 @@  static av_cold int libjxl_decode_close(AVCodecContext *avctx)
         JxlDecoderDestroy(ctx->decoder);
     ctx->decoder = NULL;
     av_buffer_unref(&ctx->iccp);
+    av_packet_free(&ctx->avpkt);
 
     return 0;
 }
@@ -468,7 +519,7 @@  const FFCodec ff_libjxl_decoder = {
     .p.id             = AV_CODEC_ID_JPEGXL,
     .priv_data_size   = sizeof(LibJxlDecodeContext),
     .init             = libjxl_decode_init,
-    FF_CODEC_DECODE_CB(libjxl_decode_frame),
+    FF_CODEC_RECEIVE_FRAME_CB(libjxl_receive_frame),
     .close            = libjxl_decode_close,
     .p.capabilities   = AV_CODEC_CAP_DR1 | AV_CODEC_CAP_OTHER_THREADS,
     .caps_internal    = FF_CODEC_CAP_NOT_INIT_THREADSAFE |
diff --git a/libavcodec/version.h b/libavcodec/version.h
index da54f87887..39dbec0208 100644
--- a/libavcodec/version.h
+++ b/libavcodec/version.h
@@ -30,7 +30,7 @@ 
 #include "version_major.h"
 
 #define LIBAVCODEC_VERSION_MINOR   6
-#define LIBAVCODEC_VERSION_MICRO 100
+#define LIBAVCODEC_VERSION_MICRO 101
 
 #define LIBAVCODEC_VERSION_INT  AV_VERSION_INT(LIBAVCODEC_VERSION_MAJOR, \
                                                LIBAVCODEC_VERSION_MINOR, \