There are seven MJPEG-tables, five small (1x12, 4x17) and two
not small (2x162). These are all avpriv, despite this not being
worthwhile due to the overhead of exporting a symbol: The total
overhead for each symbol consists of two entries in .dynsym (24B each),
one entry in the importing library's .rela.dyn (24B) and one in .got
(8B) as well as 2x2B for symbol versions and 4B for symbol hashes
in the exporting library; in addition to that, the name of the symbol
is included in both exporting and importing libraries, using 2x210 bytes
in this case.
(The above numbers are for a x64 Elf/Linux/GNU system. Other platforms
will give different numbers.)
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
The problem is that the argument 'q' is of the type uint8_t.
According to the JPEG standard, if 1 <= q <= 50, the scale factor
'S' should be 5000 / Q. Because the create_default_qtables() reuses
the variable 'q' to store the result of this calculation, for small
values of q < 19, q wil subsequently overflow and give wrong results
in the calculated quantization tables.
Instead, use a new variable 'S' (same name as in RFC2435) with the
proper range to store the result of the division.
Signed-off-by: Martin Storsjö <martin@martin.st>
Original mail and my own followup on ffmpeg-user earlier today:
I have a device sending out a MJPEG/RTP stream on a low quality setting.
Decoding and displaying the video with libavformat results in a washed
out, low contrast, greyish image. Playing the same stream with VLC results
in proper color representation.
Screenshots for comparison:
http://zevv.nl/div/libav/shot-ffplay.jpghttp://zevv.nl/div/libav/shot-vlc.jpg
A pcap capture of a few seconds of video and SDP file for playing the
stream are available at
http://zevv.nl/div/libav/mjpeg.pcaphttp://zevv.nl/div/libav/mjpeg.sdp
I believe the problem might be in the calculation of the quantization
tables in the function create_default_qtables(), the attached patch
solves the issue for me.
The problem is that the argument 'q' is of the type uint8_t. According to the
JPEG standard, if 1 <= q <= 50, the scale factor 'S' should be 5000 / Q.
Because the create_default_qtables() reuses the variable 'q' to store the
result of this calculation, for small values of q < 19, q wil subsequently
overflow and give wrong results in the calculated quantization tables. The
patch below uses a new variable 'S' (same name as in RFC2435) with the proper
range to store the result of the division.
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Many of these functions were named foo_free_context, and since
the functions no longer should free the context itself, only
allocated elements within it, the previous naming was slightly
misleading.
Signed-off-by: Martin Storsjö <martin@martin.st>
This makes it more consistent with depacketizers that don't have any
.free function at all, where the payload context is freed by the
surrounding framework. Always free the context in the surrounding
framework, having the individual depacketizers only free any data
they've specifically allocated themselves.
This is similar to how this works for demuxer/muxers/codecs - a
component shouldn't free the priv_data that the framework has
allocated for it.
Signed-off-by: Martin Storsjö <martin@martin.st>
Everything written with this bitstream writer is 8/16 bit units
(except for a pair of 4 bit values), so using a bitstream writer
isn't necessary.
Signed-off-by: Martin Storsjö <martin@martin.st>
Generalize writing of any number of qtables. Don't manually write
16 bit values in two separate calls.
Signed-off-by: Martin Storsjö <martin@martin.st>
Currently the size header of the generated DHT section is
incorrect, making the mjpeg decoder just skip it. Since the
written huffman tables are the default ones, this failure had
gone undetected.
Signed-off-by: Martin Storsjö <martin@martin.st>
Andreas Rheinhardt
Carl Eugen Hoyos
Diego Biurrun
Ico Doornekamp
Martin Storsjö
Carl Eugen Hoyos
Anton Khirnov
Samuel Pitoiset
