[FFmpeg-devel] [PATCH] png parser
Michael Niedermayer
michaelni
Thu Jun 11 13:42:32 CEST 2009
On Thu, Jun 11, 2009 at 09:35:58AM +0200, Peter Holik wrote:
> > cOn Wed, Jun 10, 2009 at 04:33:48PM +0200, Peter Holik wrote:
> >> >> On Wed, Jun 10, 2009 at 01:27:10PM +0200, Peter Holik wrote:
> >> >>> > On Wed, Jun 10, 2009 at 11:52:34AM +0200, Peter Holik wrote:
> >> >>> >> >>> On Thu, Jun 04, 2009 at 10:40:55AM +0200, Peter Holik wrote:
> >> >>> >> >>>> > On Mon, Jun 01, 2009 at 12:27:52PM +0200, Peter Holik wrote:
> >> >>> >> >>>> >> Hi!
> >> >>> >> >>>> >>
> >> >>> >> >>>> >> I've now reworked my patch to really parse a png for end of frame.
> >> >>> >> >>>> >>
> >> >>> >> >>>> >> Now you can filter out png's of any stream like
> >> >>> >> >>>> >>
> >> >>> >> >>>> >> cat garbage a.png garbage b.png garbage | ffmpeg -f image2pipe -vcodec png -i -
> >> >>> c%d.png
> >> >>> >> >>>> >>
> >> >>> >> >>>> >> cu Peter
> >> >>> >> >>>> >> Makefile | 1
> >> >>> >> >>>> >> allcodecs.c | 1
> >> >>> >> >>>> >> png_parser.c | 124 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
> >> 3
> >> >>> files
> >> >>> >> >> changed, 126 insertions(+)
> >> >>> >> >>>> >> 63f7cd1bb826726f1d2a9e7254410cd7ec98a81c png-parser.patch
> >> >>> >> >>>> >> diff --git a/libavcodec/Makefile b/libavcodec/Makefile
> >> >>> >> >>>> >> index 14ef9f4..189231b 100644
> >> >>> >> >>>> >> --- a/libavcodec/Makefile
> >> >>> >> >>>> >> +++ b/libavcodec/Makefile
> >> >>> >> >>>> >> @@ -411,6 +411,7 @@ OBJS-$(CONFIG_MLP_PARSER) += mlp_parser.o mlp.o
> >> >>> >> >>>> >> OBJS-$(CONFIG_MPEG4VIDEO_PARSER) += mpeg4video_parser.o h263.o
> >> mpeg12data.o
> >> >>> >> >>>> mpegvideo.o
> >> >>> >> >>>> >> error_resilience.o
> >> >>> >> >>>> >> OBJS-$(CONFIG_MPEGAUDIO_PARSER) += mpegaudio_parser.o
> >> mpegaudiodecheader.o
> >> >>> >> >>>> >> mpegaudiodata.o
> >> >>> >> >>>> >> OBJS-$(CONFIG_MPEGVIDEO_PARSER) += mpegvideo_parser.o mpeg12.o
> >> mpeg12data.o
> >> >>> >> >>>> mpegvideo.o
> >> >>> >> >>>> >> error_resilience.o
> >> >>> >> >>>> >> +OBJS-$(CONFIG_PNG_PARSER) += png_parser.o
> >> >>> >> >>>> >> OBJS-$(CONFIG_PNM_PARSER) += pnm_parser.o pnm.o
> >> >>> OBJS-$(CONFIG_VC1_PARSER)
> >> >>> >> >> += vc1_parser.o
> >> >>> >> >>>> >> OBJS-$(CONFIG_VP3_PARSER) += vp3_parser.o
> >> >>> >> >>>> >> diff --git a/libavcodec/allcodecs.c b/libavcodec/allcodecs.c index
> >> e8d5d27..362021f
> >> >>> >> 100644
> >> >>> >> >>>> >> --- a/libavcodec/allcodecs.c
> >> >>> >> >>>> >> +++ b/libavcodec/allcodecs.c
> >> >>> >> >>>> >> @@ -334,6 +334,7 @@ void avcodec_register_all(void)
> >> >>> >> >>>> >> REGISTER_PARSER (MPEG4VIDEO, mpeg4video);
> >> >>> >> >>>> >> REGISTER_PARSER (MPEGAUDIO, mpegaudio);
> >> >>> >> >>>> >> REGISTER_PARSER (MPEGVIDEO, mpegvideo);
> >> >>> >> >>>> >> + REGISTER_PARSER (PNG, png);
> >> >>> >> >>>> >> REGISTER_PARSER (PNM, pnm);
> >> >>> >> >>>> >> REGISTER_PARSER (VC1, vc1);
> >> >>> >> >>>> >> REGISTER_PARSER (VP3, vp3);
> >> >>> >> >>>> >> diff --git a/libavcodec/png_parser.c b/libavcodec/png_parser.c new file mode
> >> 100644
> >> >>> >> >>>> >> index 0000000..64f2356
> >> >>> >> >>>> >> --- /dev/null
> >> >>> >> >>>> >> +++ b/libavcodec/png_parser.c
> >> >>> >> >>>> >> @@ -0,0 +1,124 @@
> >> >>> >> >>>> >> +/*
> >> >>> >> >>>> >> + * PNG parser
> >> >>> >> >>>> >> + * Copyright (c) 2009 Peter Holik
> >> >>> >> >>>> >> + *
> >> >>> >> >>>> >> + * This file is part of FFmpeg.
> >> >>> >> >>>> >> + *
> >> >>> >> >>>> >> + * FFmpeg is free software; you can redistribute it and/or
> >> >>> >> >>>> >> + * modify it under the terms of the GNU Lesser General Public + * License as
> >> >>> published
> >> >>> >> by
> >> >>> >> >> the Free Software Foundation; either + * version 2.1 of the License, or (at your
> >> option)
> >> >>> any
> >> >>> >> >> later version. + *
> >> >>> >> >>>> >> + * FFmpeg is distributed in the hope that it will be useful, + * but WITHOUT ANY
> >> >>> >> WARRANTY;
> >> >>> >> >> without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR
> >> >>> >> >> PURPOSE. See the GNU + * Lesser General Public License for more details.
> >> >>> >> >>>> >> + *
> >> >>> >> >>>> >> + * You should have received a copy of the GNU Lesser General Public + * License
> >> >>> along
> >> >>> >> with
> >> >>> >> >> FFmpeg; if not, write to the Free Software + * Foundation, Inc., 51 Franklin Street,
> >> >>> Fifth
> >> >>> >> >> Floor, Boston, MA 02110-1301 USA + */
> >> >>> >> >>>> >> +
> >> >>> >> >>>> >> +/**
> >> >>> >> >>>> >> + * @file libavcodec/png_parser.c
> >> >>> >> >>>> >> + * PNG parser
> >> >>> >> >>>> >> + */
> >> >>> >> >>>> >> +
> >> >>> >> >>>> >> +#include "parser.h"
> >> >>> >> >>>> >> +#include "bytestream.h"
> >> >>> >> >>>> >> +#include "png.h"
> >> >>> >> >>>> >> +
> >> >>> >> >>>> >> +static int png_parse(AVCodecParserContext *s, AVCodecContext *avctx, +
> >> >>> >> >> const uint8_t **poutbuf, int *poutbuf_size, + const uint8_t *buf,
> >> int
> >> >>> >> >> buf_size)
> >> >>> >> >>>> >> +{
> >> >>> >> >>>> >> + ParseContext *pc = s->priv_data;
> >> >>> >> >>>> >> + uint8_t *bytestream;
> >> >>> >> >>>> >> + uint8_t *bytestream_start;
> >> >>> >> >>>> >> + uint8_t *bytestream_end;
> >> >>> >> >>>> >> +
> >> >>> >> >>>> >> + /* EOF considered as end of frame */
> >> >>> >> >>>> >> + if (buf_size == 0)
> >> >>> >> >>>> >> + goto flush;
> >> >>> >> >>>> >> +
> >> >>> >> >>>> >> + *poutbuf = NULL;
> >> >>> >> >>>> >> + *poutbuf_size = 0;
> >> >>> >> >>>> >> +
> >> >>> >> >>>> >> + /* first add new buffer to parsebuffer */
> >> >>> >> >>>> >> + bytestream_start = av_fast_realloc(pc->buffer, &pc->buffer_size, +
> >> >>> >> >> pc->index + buf_size + FF_INPUT_BUFFER_PADDING_SIZE); + if(!bytestream_start)
> >> >>> >> >>>> >> + goto fail;
> >> >>> >> >>>> >> + pc->buffer = bytestream_start;
> >> >>> >> >>>> >> + memcpy(&pc->buffer[pc->index], buf, buf_size);
> >> >>> >> >>>> >
> >> >>> >> >>>> > mandatory copying is unacceptable
> >> >>> >> >>>> > the case where each input is exactly a png frame for example should be free of any
> >> >>> >> >> timeconsuming operation
> >> >>> >> >>>> >
> >> >>> >> >>>> > maybe ff_combine_frame() would be usefull
> >> >>> >> >>>>
> >> >>> >> >>>> But ff_combine_frame also does memcpy, so there is no difference to my approach.
> >> >>> >> >>>
> >> >>> >> >>> ff_combine_frame() does not always memcpy, your code does always
> >> >>> >> >>
> >> >>> >> >> change my patch to only memcpy when input is not exactly a png frame.
> >> >>> >> >>
> >> >>> >> >> cu Peter
> >> >>> >>
> >> >>> >> ping
> >> >>> >
> >> >>> > does your code do something else than ff_combine_frame() ?
> >> >>> > if no its duplicate code
> >> >>> > if yes its either worse or ff_combine_frame() should be improved
> >> >>> >
> >> >>>
> >> >>> 1. ff_combine_frame is overhead
> >> >>
> >> >> elaborate
> >> >
> >> > there is more done than i need (pc->overread, pc->state,...)
> >> >
> >> >>> 2. i use a variable from PasteContext also altered by ff_combine_frame
> >> >>
> >> >> which variable is how used by you so that it conflicts with ff_combine_frame?
> >> >
> >> > * i use pc->last_index to remember where to parse from, because i do not like
> >> > to parse every time from the beginning.
> >> >
> >> > * to return a packet starting with a png signature i skip the garbage if any
> >> > and need this information in my combine_frame version.
> >> >
> >> >
> >> > well, i will try to use ff_combine_frame()
> >>
> >> now with ff_combine_frame()
> >
> > well yes, but ive meant it should be used like also in the other parsers,
> > but lets rather look at actual problems of your approuch ... if you can
> > fix them and your code is simpler i wouldnt mind it but i doubt it
> >
> > example:
> > 2 packets
> > packet 1: frame 0, frame 1, fra
> > packet 2: me 2, frame 3 frame 4
> >
> > your approuch will memcpy the whole 2nd packet, the normal ff_combine_frame
> > code will copy only "frame 2"
>
> ff_combine_frame copy as much as you tell (ff_combine_frame don't know where frame 2 ends)
>
>
> you are right, but i don't think that this situation with your 2 packets ever happens.
>
> usually there is the opposite situation:
>
> packet 1: frame 0(part 1)
> packet 2: frame 0(last part), frame 1(part 1)
>
> because image2pipe uses only a 4kByte Buffer.
thats all fine as long as the parser is just used with image2pipe and
image2pipe buffer is not changed ...
i think these are not reasonable assumtions
>
>
> My main problem with png is, that i never know the total size of the picture.
> I read chunks (with a length)
>
> And most troubles are if for example the length (32bits) of the chunk-header
> is across two packets so i decided to read/copy ahead (of course copy more than i really would need).
some rough idea:
uint64_t state= state_from_context;
for(i=0; i<buf_size; i++){
state= (state<<8) + buf[i];
if(state == startcode){
we have a new png start
}
}
for(; i<buf_size; i++){
int skip= FFMIN(remaining-8, buf_size-i-1);
if(skip>0){
i += skip;
remaining -= skip;
}
state= (state<<8) + buf[i];
remaining--;
if(remaining == 0){
remaining= state&0xFFFFFFFF;
if(state>>32 == end)
png end
}
}
[...]
--
Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB
In fact, the RIAA has been known to suggest that students drop out
of college or go to community college in order to be able to afford
settlements. -- The RIAA
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
URL: <http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/attachments/20090611/198c00ff/attachment.pgp>
More information about the ffmpeg-devel
mailing list