gmane.comp.video.ffmpeg.devel

Re: [PATCH]Fix for issue694. Dirac A/V sync loss

Anuradha Suraparaju — 2008-12-02T02:39:03

Hi, Sorry for the delayed response. My replies are inline. On Sat, 2008-11-29 at 02:44 +0100, Michael Niedermayer wrote: Dirac, the specification, has a flexible GOP structure. So the frame re-ordering can be anything. This said, the current implementations of Dirac (dirac-research and Schroedinger) use a frame-reordering on 1. Intra and backward predicted frames can be delayed but bi-directional frames are not. So the mpeg1/2 logic for has_b_frames holds. In Dirac it is not possible to tell whether a frame is a backward predicted frame (similar to 'P' frame) or a bi-directional frame (similar to 'B' frame) easily without processing its reference frames. So I set has_b_frames to 1 if I come across any predicted frame (buf[4] &0x08) == 0x08 - mean the parse unit is a picture, and (buf[4] & 0x03 checks the number of references) . So an I-frame only sequence will set has_b_frames to 0 but a sequence having only P or P&B frames will set it to 1 since has_b_frames=1 does not require that there be any 'B' frames in the sequence. Will change it to av_fast_realloc in the next update to the patch. Will fix this. Dirac is a bit trickier. Since we use Arithmetic coding for entropy coding the sync sequence 'BBCD' can occur in the entropy data and can trigger a false start of frame. We can't go by frame size (here the next parse offset) alone because of this. So we need to use the parse offsets of the current frame and the next frame to validate the current frame. The previous parse offset of the next frame must match the next parse offset of the current frame. If they don't match the current frame gets thrown out. So I decided to implement a Dirac specific combine frame. Hope this explains it. Regards, Anuradha

Re: [PATCH] QCELP decoder

Kenan Gillet — 2008-12-02T01:45:26

Hi, On Mon, Dec 1, 2008 at 3:32 PM, Kenan Gillet gmail.com> wrote: here is an updated version of qcelp_lsp.c which had been already oked Index: libavcodec/qcelp_lsp.c =================================================================== --- libavcodec/qcelp_lsp.c(revision 0) +++ libavcodec/qcelp_lsp.c(revision 0) < at >< at > -0,0 +1,100 < at >< at > +/* + * QCELP decoder + * Copyright (c) 2007 Reynaldo H. Verdejo Pinochet + * + * 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 + */ + +/** + * < at >file qcelp_lsp.c + * QCELP decoder + * < at >author Reynaldo H. Verdejo Pinochet + */ + +#include "libavutil/mathematics.h" + +/** + * initial coefficient to perform bandwidth expansion on LPC + * + * < at >note: 0.9883 looks like an approximation of 253/256. + * + * TIA/EIA/IS-733 2.4.3.3.6 6 + */ +#define QCELP_BANDWITH_EXPANSION_COEFF 0.9883 + +/** + * Computes the Pa / (1 + z(-1)) or Qa / (1 - z(-1)) coefficients + * needed for LSP to LPC conversion. + * We only need to calculate the 6 first elements of the polynomial. + * + * < at >param lspf line spectral pair frequencies + * < at >param f [out] polynomial input/output as a vector + * + * TIA/EIA/IS-733 2.4.3.3.5-1/2 + */ +static void lsp2polyf(const float *lspf, double *f, int lp_half_order) +{ + int i, j; + + f[0] = 1.0; + f[1] = -2 * cos(M_PI * lspf[0]); + lspf -= 2; + for(i=2; i<=lp_half_order; i++) + { + double val = -2 * cos(M_PI * lspf[2*i]); + f[i] = val * f[i-1] + 2*f[i-2]; + for(j=i-1; j>1; j--) + f[j] += f[j-1] * val + f[j-2]; + f[1] += val; + } +} + +/** + * Reconstructs LPC coefficients from the line spectral pair frequencies + * and performs bandwidth expansion. + * + * < at >param lspf line spectral pair frequencies + * < at >param lpc linear predictive coding coefficients + * + * < at >note: bandwith_expansion_coeff could be precalculated into a table + * but it seems to be slower on x86 + * + * TIA/EIA/IS-733 2.4.3.3.5 + */ +void ff_qcelp_lspf2lpc(const float *lspf, float *lpc) +{ + double pa[6], qa[6]; + int i; + double bandwith_expansion_coeff = -QCELP_BANDWITH_EXPANSION_COEFF * 0.5; + + lsp2polyf(lspf, pa, 5); + lsp2polyf(lspf + 1, qa, 5); + + for (i=4; i>=0; i--) + { + double paf = pa[i+1] + pa[i]; + double qaf = qa[i+1] - qa[i]; + + lpc[i ] = paf + qaf; + lpc[9-i] = paf - qaf; + } + for (i=0; i<10; i++) + { + lpc[i] *= bandwith_expansion_coeff; + bandwith_expansion_coeff *= QCELP_BANDWITH_EXPANSION_COEFF; + } +} _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel< at >mplayerhq.hu https://lists.mplayerhq.hu/mailman/listinfo/ffmpeg-devel

Re: [PATCH] remove useless extern keywords

Måns Rullgård — 2008-12-02T01:32:36

I'll comment on the files I maintain. Wrong. Those are function pointers. OK OK. No maintainer, but I seem to be the only one around here who knows SH4. (Yes, those should have an ff_ prefix.) OK OK OK OK OK OK Those declarations don't belong there at all. The definitions are in utils.c. I don't know what the proper place would be. Michael?

[PATCH] remove useless extern keywords

Diego Biurrun — 2008-12-02T01:16:03

In the nvidia patch review there were complaints about unnecessary extern keywords being added to the code. Since this was apparently a copy and paste issue from the xvmc code, I wondered why they were there in the first place. Here is a patch to get rid of them so no bad examples remain. Diego _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel< at >mplayerhq.hu https://lists.mplayerhq.hu/mailman/listinfo/ffmpeg-devel

Re: [PATCH] remove duplicate declarationsfromlibswscale/rgb2rgb.c

Diego Biurrun — 2008-12-02T01:13:22

Forget this one. Diego

[PATCH] remove duplicate declarations fromlibswscale/rgb2rgb.c

Diego Biurrun — 2008-12-02T00:32:08

There are a bunch of pointless forward declarations in rgb2rgb.c. The same already exist in rgb2rgb.h. Here is a patch to remove them. Diego _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel< at >mplayerhq.hu https://lists.mplayerhq.hu/mailman/listinfo/ffmpeg-devel

Re: [PATCH] remove unused codefromlibswscale/rgb2rgb.c

Diego Biurrun — 2008-12-02T00:22:00

Applied. Diego

Re: [PATCH] remove unused codefromlibswscale/rgb2rgb.c

Michael Niedermayer — 2008-12-02T00:16:08

_______________________________________________ ffmpeg-devel mailing list ffmpeg-devel< at >mplayerhq.hu https://lists.mplayerhq.hu/mailman/listinfo/ffmpeg-devel

[PATCH] remove unused code from libswscale/rgb2rgb.c

Diego Biurrun — 2008-12-02T00:00:01

Here is a patch to remove a bunch of unused declarations from libswscale/rgb2rgb.c. Diego _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel< at >mplayerhq.hu https://lists.mplayerhq.hu/mailman/listinfo/ffmpeg-devel

Re: [PATCH] QCELP decoder

Michael Niedermayer — 2008-12-01T23:57:11

_______________________________________________ ffmpeg-devel mailing list ffmpeg-devel< at >mplayerhq.hu https://lists.mplayerhq.hu/mailman/listinfo/ffmpeg-devel

Re: [PATCH]NVIDIA VDPAU patch for h264

Jason Garrett-Glaser — 2008-12-01T23:52:54

I have yet to see anyone use Baseline features that aren't a subset of Main for any serious purpose. Thus, your approach of splitting things by profiles and "only playing Main and High" will completely kill support for all Baseline streams which would have otherwise worked just fine. I have never seen FMO, ASO, or RS used outside of a research paper. Splitting by profile simply doesn't make any sense to me. If you need to pass the profile to the decoder, why not just create a variable and enums? Certainly not, since I doubt your graphics cards support SVQ3! SVQ3 is a proprietary ripoff of H.264 with some minor modifications used by some Quicktime videos. RV30/RV40 are similar, except ripoffs by Real instead of Sorenson. I'm pretty sure you should go back and read the C spec here. I don't think we should let features that nobody use and libavcodec doesn't (and probably will never) support hold us up for implementing things the right way. Dark Shikari

Re: [PATCH] QCELP decoder

Kenan Gillet — 2008-12-01T23:32:00

On Dec 1, 2008, at 3:18 PM, Michael Niedermayer wrote: how about a note in the doxy? < at >note: 0.9883 looks like an approximation of 253/256 I'll add it to the next round of patches if it is ok. :)

Re: [PATCH] QCELP decoder

Michael Niedermayer — 2008-12-01T23:25:15

_______________________________________________ ffmpeg-devel mailing list ffmpeg-devel< at >mplayerhq.hu https://lists.mplayerhq.hu/mailman/listinfo/ffmpeg-devel

Re: [PATCH] QCELP decoder

Michael Niedermayer — 2008-12-01T23:18:11

_______________________________________________ ffmpeg-devel mailing list ffmpeg-devel< at >mplayerhq.hu https://lists.mplayerhq.hu/mailman/listinfo/ffmpeg-devel

Re: [PATCH] QCELP decoder

Kenan Gillet — 2008-12-01T23:15:40

Hi, On Mon, Dec 1, 2008 at 1:39 PM, Michael Niedermayer gmx.at> wrote: [...] [...] done, and var renamed to pitch_lag done, and var renamed to pitch_gain [...] when removed, some samples wav output changes, plus QCELP only handles mono. round 15 attached. note: there is a small update to the glue code in libavformat/mov.c so I repost the doc-glue patch for this particular round but it has not changed otherwise. thanks for all your help, Kenan Index: libavcodec/qcelpdec.c =================================================================== --- libavcodec/qcelpdec.c(revision 15975) +++ libavcodec/qcelpdec.c(working copy) < at >< at > -65,8 +65,8 < at >< at > float last_codebook_gain; int prev_g1[2]; int prev_bitrate; - float prev_pitch_gain[4]; - uint8_t prev_pitch_lag[4]; + float pitch_gain[4]; + uint8_t pitch_lag[4]; uint16_t first16bits; } QCELPContext; < at >< at > -75,7 +75,7 < at >< at > * * TIA/EIA/IS-733 2.4.3.3.5 */ -void qcelp_lspf2lpc(const float *lspf, float *lpc); +void ff_qcelp_lspf2lpc(const float *lspf, float *lpc); static void weighted_vector_sumf(float *out, const float *in_a, const float *in_b, float weight_coeff_a, < at >< at > -495,6 +495,58 < at >< at > } /** + * Apply pitch synthesis filter and pitch prefilter to the scaled codebook vector. + * TIA/EIA/IS-733 2.4.5.2 + * + * < at >param q the context + * < at >param cdn_vector the scaled codebook vector + */ +static void apply_pitch_filters(QCELPContext *q, + float *cdn_vector) { + int i; + const float *v_synthesis_filtered, *v_pre_filtered; + + if (q->bitrate >= RATE_HALF || + (q->bitrate == I_F_Q && (q->prev_bitrate >= RATE_HALF))) { + + if (q->bitrate >= RATE_HALF) { + + // Compute gain & lag for the whole frame. + for (i = 0; i < 4; i++) { + q->pitch_gain[i] = q->frame.plag[i] ? (q->frame.pgain[i] + 1) * 0.25 : 0.0; + + q->pitch_lag[i] = q->frame.plag[i] + 16; + } + } else { + float max_pitch_gain = q->erasure_count < 3 ? 0.9 - 0.3 * (q->erasure_count - 1) + : 0.0; + for (i = 0; i < 4; i++) + q->pitch_gain[i] = FFMIN(q->pitch_gain[i], max_pitch_gain); + + memset(q->frame.pfrac, 0, sizeof(q->frame.pfrac)); + } + + // pitch synthesis filter + v_synthesis_filtered = do_pitchfilter(q->pitch_synthesis_filter_mem, cdn_vector, + q->pitch_gain, q->pitch_lag, q->frame.pfrac); + + // pitch prefilter update + for (i = 0; i < 4; i++) + q->pitch_gain[i] = 0.5 * FFMIN(q->pitch_gain[i], 1.0); + + v_pre_filtered = do_pitchfilter(q->pitch_pre_filter_mem, v_synthesis_filtered, + q->pitch_gain, q->pitch_lag, q->frame.pfrac); + + apply_gain_ctrl(cdn_vector, v_synthesis_filtered, v_pre_filtered); + } else { + memcpy(q->pitch_synthesis_filter_mem, cdn_vector + 17, 143 * sizeof(float)); + memcpy(q->pitch_pre_filter_mem, cdn_vector + 17, 143 * sizeof(float)); + memset(q->pitch_gain, 0, sizeof(q->pitch_gain)); + memset(q->pitch_lag, 0, sizeof(q->pitch_lag)); + } +} + +/** * Interpolates LSP frequencies and computes LPC coefficients * for a given bitrate & pitch subframe. * < at >< at > -522,9 +574,9 < at >< at > { weighted_vector_sumf(interpolated_lspf, curr_lspf, q->prev_lspf, weight, 1.0 - weight, 10); - qcelp_lspf2lpc(interpolated_lspf, lpc); + ff_qcelp_lspf2lpc(interpolated_lspf, lpc); }else if(q->bitrate >= RATE_QUARTER || (q->bitrate == I_F_Q && !subframe_num)) - qcelp_lspf2lpc(curr_lspf, lpc); + ff_qcelp_lspf2lpc(curr_lspf, lpc); } static int buf_size2bitrate(const int buf_size) Index: Changelog =================================================================== --- Changelog(revision 15975) +++ Changelog(working copy) < at >< at > -141,6 +141,7 < at >< at > - Speex decoding via libspeex - Electronic Arts TGQ decoder - RV40 decoder +- QCELP / PureVoice decoder version 0.4.9-pre1: Index: libavcodec/Makefile =================================================================== --- libavcodec/Makefile(revision 15975) +++ libavcodec/Makefile(working copy) < at >< at > -158,6 +158,7 < at >< at > OBJS-$(CONFIG_PNG_ENCODER) += png.o pngenc.o OBJS-$(CONFIG_PPM_ENCODER) += pnmenc.o pnm.o OBJS-$(CONFIG_PTX_DECODER) += ptx.o +OBJS-$(CONFIG_QCELP_DECODER) += qcelpdec.o qcelp_lsp.o celp_math.o celp_filters.o OBJS-$(CONFIG_QDM2_DECODER) += qdm2.o mpegaudiodec.o mpegaudiodecheader.o mpegaudio.o mpegaudiodata.o OBJS-$(CONFIG_QDRAW_DECODER) += qdrw.o OBJS-$(CONFIG_QPEG_DECODER) += qpeg.o Index: libavcodec/allcodecs.c =================================================================== --- libavcodec/allcodecs.c(revision 15975) +++ libavcodec/allcodecs.c(working copy) < at >< at > -204,6 +204,7 < at >< at > REGISTER_DECODER (MPC7, mpc7); REGISTER_DECODER (MPC8, mpc8); REGISTER_ENCDEC (NELLYMOSER, nellymoser); + REGISTER_DECODER (QCELP, qcelp); REGISTER_DECODER (QDM2, qdm2); REGISTER_DECODER (RA_144, ra_144); REGISTER_DECODER (RA_288, ra_288); Index: libavcodec/avcodec.h =================================================================== --- libavcodec/avcodec.h(revision 15975) +++ libavcodec/avcodec.h(working copy) < at >< at > -30,7 +30,7 < at >< at > #include "libavutil/avutil.h" #define LIBAVCODEC_VERSION_MAJOR 52 -#define LIBAVCODEC_VERSION_MINOR 5 +#define LIBAVCODEC_VERSION_MINOR 6 #define LIBAVCODEC_VERSION_MICRO 0 #define LIBAVCODEC_VERSION_INT AV_VERSION_INT(LIBAVCODEC_VERSION_MAJOR, \ Index: doc/general.texi =================================================================== --- doc/general.texi(revision 15975) +++ doc/general.texi(working copy) < at >< at > -397,6 +397,7 < at >< at > < at >item Musepack < at >tab < at >tab X < at >tab SV7 and SV8 are supported. < at >item Nellymoser ASAO < at >tab X < at >tab X +< at >item QCELP / PureVoice < at >tab < at >tab X < at >item Qdesign QDM2 < at >tab < at >tab X < at >tab There are still some distortions. < at >item QT IMA ADPCM < at >tab X < at >tab X Index: libavformat/mov.c =================================================================== --- libavformat/mov.c(revision 15975) +++ libavformat/mov.c(working copy) < at >< at > -988,7 +988,9 < at >< at > #endif /* no ifdef since parameters are always those */ case CODEC_ID_QCELP: + st->codec->frame_size= 160; st->codec->channels= 1; /* really needed */ + break; case CODEC_ID_AMR_NB: case CODEC_ID_AMR_WB: st->codec->frame_size= sc->samples_per_frame; Index: libavcodec/qcelp_lsp.c =================================================================== --- libavcodec/qcelp_lsp.c(revision 0) +++ libavcodec/qcelp_lsp.c(revision 0) < at >< at > -0,0 +1,98 < at >< at > +/* + * QCELP decoder + * Copyright (c) 2007 Reynaldo H. Verdejo Pinochet + * + * 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 + */ + +/** + * < at >file qcelp_lsp.c + * QCELP decoder + * < at >author Reynaldo H. Verdejo Pinochet + */ + +#include "libavutil/mathematics.h" + +/** + * initial coefficient to perform bandwidth expansion on LPC + * + * TIA/EIA/IS-733 2.4.3.3.6 6 + */ +#define QCELP_BANDWITH_EXPANSION_COEFF 0.9883 + +/** + * Computes the Pa / (1 + z(-1)) or Qa / (1 - z(-1)) coefficients + * needed for LSP to LPC conversion. + * We only need to calculate the 6 first elements of the polynomial. + * + * < at >param lspf line spectral pair frequencies + * < at >param f [out] polynomial input/output as a vector + * + * TIA/EIA/IS-733 2.4.3.3.5-1/2 + */ +static void lsp2polyf(const float *lspf, double *f, int lp_half_order) +{ + int i, j; + + f[0] = 1.0; + f[1] = -2 * cos(M_PI * lspf[0]); + lspf -= 2; + for(i=2; i<=lp_half_order; i++) + { + double val = -2 * cos(M_PI * lspf[2*i]); + f[i] = val * f[i-1] + 2*f[i-2]; + for(j=i-1; j>1; j--) + f[j] += f[j-1] * val + f[j-2]; + f[1] += val; + } +} + +/** + * Reconstructs LPC coefficients from the line spectral pair frequencies + * and performs bandwidth expansion. + * + * < at >param lspf line spectral pair frequencies + * < at >param lpc linear predictive coding coefficients + * + * < at >note: bandwith_expansion_coeff could be precalculated into a table + * but it seems to be slower on x86 + * + * TIA/EIA/IS-733 2.4.3.3.5 + */ +void ff_qcelp_lspf2lpc(const float *lspf, float *lpc) +{ + double pa[6], qa[6]; + int i; + double bandwith_expansion_coeff = -QCELP_BANDWITH_EXPANSION_COEFF * 0.5; + + lsp2polyf(lspf, pa, 5); + lsp2polyf(lspf + 1, qa, 5); + + for (i=4; i>=0; i--) + { + double paf = pa[i+1] + pa[i]; + double qaf = qa[i+1] - qa[i]; + + lpc[i ] = paf + qaf; + lpc[9-i] = paf - qaf; + } + for (i=0; i<10; i++) + { + lpc[i] *= bandwith_expansion_coeff; + bandwith_expansion_coeff *= QCELP_BANDWITH_EXPANSION_COEFF; + } +} _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel< at >mplayerhq.hu https://lists.mplayerhq.hu/mailman/listinfo/ffmpeg-devel

Re: [PATCH] QCELP decoder

Kenan Gillet — 2008-12-01T23:00:49

On Dec 1, 2008, at 2:05 PM, Michael Niedermayer wrote: probably but in specs and reference code it is 0.9883 not 0.98828125. do you want to replace 0.9883 by 253/256 ? done done also I am interested to know if u had any comments on the benchmarks [1] between old double vs old float and new double vs new float [1] http://lists.mplayerhq.hu/pipermail/ffmpeg-devel/2008-December/056795.html

Re: [PATCH] QCELP decoder

Vitor Sessak — 2008-12-01T22:07:46

Ok'ed parts applied. -Vitor

Re: [PATCH] QCELP decoder

Michael Niedermayer — 2008-12-01T22:05:18

_______________________________________________ ffmpeg-devel mailing list ffmpeg-devel< at >mplayerhq.hu https://lists.mplayerhq.hu/mailman/listinfo/ffmpeg-devel

Re: [PATCH]NVIDIA VDPAU patch for h264

compn — 2008-12-01T21:51:38

you can check out some ffmpeg guidelines here: http://ffmpeg.org/general.html#SEC21 -compn

Re: [PATCH] QCELP decoder

Diego Biurrun — 2008-12-01T21:48:42

The glue/documentation/build system patch is OK, no need to resend it. Diego

Re: [PATCH] QCELP decoder

Michael Niedermayer — 2008-12-01T21:41:47

_______________________________________________ ffmpeg-devel mailing list ffmpeg-devel< at >mplayerhq.hu https://lists.mplayerhq.hu/mailman/listinfo/ffmpeg-devel