Merge branch 'master' of git://git.videolan.org/ffmpeg
authorMichael Bradshaw <mbradshaw@sorensonmedia.com>
Fri, 18 Nov 2011 21:04:11 +0000 (14:04 -0700)
committerMichael Bradshaw <mbradshaw@sorensonmedia.com>
Fri, 18 Nov 2011 21:04:11 +0000 (14:04 -0700)
Changelog
doc/filters.texi
libavcodec/h264.c
libavcodec/qdm2.c
libavfilter/Makefile
libavfilter/af_pan.c [new file with mode: 0644]
libavfilter/allfilters.c
libavfilter/avfilter.h

index 0124479..19472d4 100644 (file)
--- a/Changelog
+++ b/Changelog
@@ -122,6 +122,7 @@ easier to use. The changes are:
 - VBLE Decoder
 - OS X Video Decoder Acceleration (VDA) support
 - compact and csv output in ffprobe
+- pan audio filter
 
 
 version 0.8:
index 9fa9c04..cefb8ad 100644 (file)
@@ -235,6 +235,54 @@ the listener (standard for speakers).
 
 Ported from SoX.
 
+@section pan
+
+Mix channels with specific gain levels. The filter accepts the output
+channel layout followed by a set of channels definitions.
+
+The filter accepts parameters of the form:
+"@var{l}:@var{outdef}:@var{outdef}:..."
+
+@table @option
+@item l
+output channel layout or number of channels
+
+@item outdef
+output channel specification, of the form:
+"@var{out_name}=[@var{gain}*]@var{in_name}[+[@var{gain}*]@var{in_name}...]"
+
+@item out_name
+output channel to define, either a channel name (FL, FR, etc.) or a channel
+number (c0, c1, etc.)
+
+@item gain
+multiplicative coefficient for the channel, 1 leaving the volume unchanged
+
+@item in_name
+input channel to use, see out_name for details; it is not possible to mix
+named and numbered input channels
+@end table
+
+If the `=' in a channel specification is replaced by `<', then the gains for
+that specification will be renormalized so that the total is 1, thus
+avoiding clipping noise.
+
+For example, if you want to down-mix from stereo to mono, but with a bigger
+factor for the left channel:
+@example
+pan=1:c0=0.9*c0+0.1*c1
+@end example
+
+A customized down-mix to stereo that works automatically for 3-, 4-, 5- and
+7-channels surround:
+@example
+pan=stereo: FL < FL + 0.5*FC + 0.6*BL + 0.6*SL : FR < FR + 0.5*FC + 0.6*BR + 0.6*SR
+@end example
+
+Note that @file{ffmpeg} integrates a default down-mix (and up-mix) system
+that should be preferred (see "-ac" option) unless you have very specific
+needs.
+
 @section volume
 
 Adjust the input audio volume.
index c389f96..b659c38 100644 (file)
@@ -2367,7 +2367,7 @@ static void implicit_weight_table(H264Context *h, int field){
 static void idr(H264Context *h){
     int i;
     ff_h264_remove_all_refs(h);
-    h->prev_frame_num= 0;
+    h->prev_frame_num= -1;
     h->prev_frame_num_offset= 0;
     h->prev_poc_msb=
     h->prev_poc_lsb= 0;
@@ -2882,7 +2882,7 @@ static int decode_slice_header(H264Context *h, H264Context *h0){
 
     if(h0->current_slice == 0){
         // Shorten frame num gaps so we don't have to allocate reference frames just to throw them away
-        if(h->frame_num != h->prev_frame_num) {
+        if(h->frame_num != h->prev_frame_num && h->prev_frame_num >= 0) {
             int unwrap_prev_frame_num = h->prev_frame_num, max_frame_num = 1<<h->sps.log2_max_frame_num;
 
             if (unwrap_prev_frame_num > h->frame_num) unwrap_prev_frame_num -= max_frame_num;
@@ -2896,7 +2896,7 @@ static int decode_slice_header(H264Context *h, H264Context *h0){
             }
         }
 
-        while(h->frame_num !=  h->prev_frame_num &&
+        while(h->frame_num !=  h->prev_frame_num && h->prev_frame_num >= 0 &&
               h->frame_num != (h->prev_frame_num+1)%(1<<h->sps.log2_max_frame_num)){
             Picture *prev = h->short_ref_count ? h->short_ref[0] : NULL;
             av_log(h->s.avctx, AV_LOG_DEBUG, "Frame num gap %d %d\n", h->frame_num, h->prev_frame_num);
index fe785af..a057c1e 100644 (file)
@@ -1329,7 +1329,7 @@ static void qdm2_fft_decode_tones (QDM2Context *q, int duration, GetBitContext *
     local_int_10 = 1 << (q->group_order - duration - 1);
     offset = 1;
 
-    while (1) {
+    while (get_bits_left(gb)>0) {
         if (q->superblocktype_2_3) {
             while ((n = qdm2_get_vlc(gb, &vlc_tab_fft_tone_offset[local_int_8], 1, 2)) < 2) {
                 offset = 1;
index 8e43be8..cab6f2e 100644 (file)
@@ -29,6 +29,7 @@ OBJS-$(CONFIG_ANULL_FILTER)                  += af_anull.o
 OBJS-$(CONFIG_ARESAMPLE_FILTER)              += af_aresample.o
 OBJS-$(CONFIG_ASHOWINFO_FILTER)              += af_ashowinfo.o
 OBJS-$(CONFIG_EARWAX_FILTER)                 += af_earwax.o
+OBJS-$(CONFIG_PAN_FILTER)                    += af_pan.o
 OBJS-$(CONFIG_VOLUME_FILTER)                 += af_volume.o
 
 OBJS-$(CONFIG_ABUFFER_FILTER)                += asrc_abuffer.o
diff --git a/libavfilter/af_pan.c b/libavfilter/af_pan.c
new file mode 100644 (file)
index 0000000..c4e64c8
--- /dev/null
@@ -0,0 +1,306 @@
+/*
+ * Copyright (c) 2002 Anders Johansson <ajh@atri.curtin.edu.au>
+ * Copyright (c) 2011 Clément Bœsch <ubitux@gmail.com>
+ * Copyright (c) 2011 Nicolas George <nicolas.george@normalesup.org>
+ *
+ * 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 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
+ * Audio panning filter (channels mixing)
+ * Original code written by Anders Johansson for MPlayer,
+ * reimplemented for FFmpeg.
+ */
+
+#include <stdio.h>
+#include "libavutil/audioconvert.h"
+#include "libavutil/avstring.h"
+#include "avfilter.h"
+#include "internal.h"
+
+#define MAX_CHANNELS 63
+
+typedef struct {
+    int64_t out_channel_layout;
+    union {
+        double d[MAX_CHANNELS][MAX_CHANNELS];
+        // i is 1:7:8 fixed-point, i.e. in [-128*256; +128*256[
+        int    i[MAX_CHANNELS][MAX_CHANNELS];
+    } gain;
+    int64_t need_renorm;
+    int need_renumber;
+    int nb_input_channels;
+    int nb_output_channels;
+} PanContext;
+
+static int parse_channel_name(char **arg, int *rchannel, int *rnamed)
+{
+    char buf[8];
+    int len, i, channel_id;
+    int64_t layout, layout0;
+
+    if (sscanf(*arg, " %7[A-Z] %n", buf, &len)) {
+        layout0 = layout = av_get_channel_layout(buf);
+        for (i = 32; i > 0; i >>= 1) {
+            if (layout >= (int64_t)1 << i) {
+                channel_id += i;
+                layout >>= i;
+            }
+        }
+        if (channel_id >= MAX_CHANNELS || layout0 != (int64_t)1 << channel_id)
+            return AVERROR(EINVAL);
+        *rchannel = channel_id;
+        *rnamed = 1;
+        *arg += len;
+        return 0;
+    }
+    if (sscanf(*arg, " c%d %n", &channel_id, &len) &&
+        channel_id >= 0 && channel_id < MAX_CHANNELS) {
+        *rchannel = channel_id;
+        *rnamed = 0;
+        *arg += len;
+        return 0;
+    }
+    return AVERROR(EINVAL);
+}
+
+static void skip_spaces(char **arg)
+{
+    int len = 0;
+
+    sscanf(*arg, " %n", &len);
+    *arg += len;
+}
+
+static av_cold int init(AVFilterContext *ctx, const char *args0, void *opaque)
+{
+    PanContext *const pan = ctx->priv;
+    char *arg, *arg0, *tokenizer, *args = av_strdup(args0);
+    int out_ch_id, in_ch_id, len, named;
+    int nb_in_channels[2] = { 0, 0 }; // number of unnamed and named input channels
+    double gain;
+
+    if (!args)
+        return AVERROR(ENOMEM);
+    arg = av_strtok(args, ":", &tokenizer);
+    pan->out_channel_layout = av_get_channel_layout(arg);
+    if (!pan->out_channel_layout) {
+        av_log(ctx, AV_LOG_ERROR, "Unknown channel layout \"%s\"\n", arg);
+        return AVERROR(EINVAL);
+    }
+    pan->nb_output_channels = av_get_channel_layout_nb_channels(pan->out_channel_layout);
+
+    /* parse channel specifications */
+    while ((arg = arg0 = av_strtok(NULL, ":", &tokenizer))) {
+        /* channel name */
+        if (parse_channel_name(&arg, &out_ch_id, &named)) {
+            av_log(ctx, AV_LOG_ERROR,
+                   "Expected out channel name, got \"%.8s\"\n", arg);
+            return AVERROR(EINVAL);
+        }
+        if (named) {
+            if (!((pan->out_channel_layout >> out_ch_id) & 1)) {
+                av_log(ctx, AV_LOG_ERROR,
+                       "Channel \"%.8s\" does not exist in the chosen layout\n", arg0);
+                return AVERROR(EINVAL);
+            }
+            /* get the channel number in the output channel layout:
+             * out_channel_layout & ((1 << out_ch_id) - 1) are all the
+             * channels that come before out_ch_id,
+             * so their count is the index of out_ch_id */
+            out_ch_id = av_get_channel_layout_nb_channels(pan->out_channel_layout & (((int64_t)1 << out_ch_id) - 1));
+        }
+        if (out_ch_id < 0 || out_ch_id >= pan->nb_output_channels) {
+            av_log(ctx, AV_LOG_ERROR,
+                   "Invalid out channel name \"%.8s\"\n", arg0);
+            return AVERROR(EINVAL);
+        }
+        if (*arg == '=') {
+            arg++;
+        } else if (*arg == '<') {
+            pan->need_renorm |= (int64_t)1 << out_ch_id;
+            arg++;
+        } else {
+            av_log(ctx, AV_LOG_ERROR,
+                   "Syntax error after channel name in \"%.8s\"\n", arg0);
+            return AVERROR(EINVAL);
+        }
+        /* gains */
+        while (1) {
+            gain = 1;
+            if (sscanf(arg, " %lf %n* %n", &gain, &len, &len))
+                arg += len;
+            if (parse_channel_name(&arg, &in_ch_id, &named)){
+                av_log(ctx, AV_LOG_ERROR,
+                       "Expected in channel name, got \"%.8s\"\n", arg);
+                return AVERROR(EINVAL);
+            }
+            nb_in_channels[named]++;
+            if (nb_in_channels[!named]) {
+                av_log(ctx, AV_LOG_ERROR,
+                       "Can not mix named and numbered channels\n");
+                return AVERROR(EINVAL);
+            }
+            pan->gain.d[out_ch_id][in_ch_id] = gain;
+            if (!*arg)
+                break;
+            if (*arg != '+') {
+                av_log(ctx, AV_LOG_ERROR, "Syntax error near \"%.8s\"\n", arg);
+                return AVERROR(EINVAL);
+            }
+            arg++;
+            skip_spaces(&arg);
+        }
+    }
+    pan->need_renumber = !!nb_in_channels[1];
+
+    av_free(args);
+    return 0;
+}
+
+static int query_formats(AVFilterContext *ctx)
+{
+    PanContext *pan = ctx->priv;
+    AVFilterLink *inlink  = ctx->inputs[0];
+    AVFilterLink *outlink = ctx->outputs[0];
+    AVFilterFormats *formats;
+
+    const enum AVSampleFormat sample_fmts[] = {AV_SAMPLE_FMT_S16, -1};
+    const int                packing_fmts[] = {AVFILTER_PACKED,   -1};
+
+    avfilter_set_common_sample_formats (ctx, avfilter_make_format_list(sample_fmts));
+    avfilter_set_common_packing_formats(ctx, avfilter_make_format_list(packing_fmts));
+
+    // inlink supports any channel layout
+    formats = avfilter_make_all_channel_layouts();
+    avfilter_formats_ref(formats, &inlink->out_chlayouts);
+
+    // outlink supports only requested output channel layout
+    formats = NULL;
+    avfilter_add_format(&formats, pan->out_channel_layout);
+    avfilter_formats_ref(formats, &outlink->in_chlayouts);
+    return 0;
+}
+
+static int config_props(AVFilterLink *link)
+{
+    AVFilterContext *ctx = link->dst;
+    PanContext *pan = ctx->priv;
+    char buf[1024], *cur;
+    int i, j, k, r;
+    double t;
+
+    pan->nb_input_channels = av_get_channel_layout_nb_channels(link->channel_layout);
+    if (pan->need_renumber) {
+        // input channels were given by their name: renumber them
+        for (i = j = 0; i < MAX_CHANNELS; i++) {
+            if ((link->channel_layout >> i) & 1) {
+                for (k = 0; k < pan->nb_output_channels; k++)
+                    pan->gain.d[k][j] = pan->gain.d[k][i];
+                j++;
+            }
+        }
+    }
+    // renormalize
+    for (i = 0; i < pan->nb_output_channels; i++) {
+        if (!((pan->need_renorm >> i) & 1))
+            continue;
+        t = 0;
+        for (j = 0; j < pan->nb_input_channels; j++)
+            t += pan->gain.d[i][j];
+        if (t > -1E-5 && t < 1E-5) {
+            // t is almost 0 but not exactly, this is probably a mistake
+            if (t)
+                av_log(ctx, AV_LOG_WARNING,
+                       "Degenerate coefficients while renormalizing\n");
+            continue;
+        }
+        for (j = 0; j < pan->nb_input_channels; j++)
+            pan->gain.d[i][j] /= t;
+    }
+    // summary
+    for (i = 0; i < pan->nb_output_channels; i++) {
+        cur = buf;
+        for (j = 0; j < pan->nb_input_channels; j++) {
+            r = snprintf(cur, buf + sizeof(buf) - cur, "%s%.3g i%d",
+                         j ? " + " : "", pan->gain.d[i][j], j);
+            cur += FFMIN(buf + sizeof(buf) - cur, r);
+        }
+        av_log(ctx, AV_LOG_INFO, "o%d = %s\n", i, buf);
+    }
+    // convert to integer
+    for (i = 0; i < pan->nb_output_channels; i++) {
+        for (j = 0; j < pan->nb_input_channels; j++) {
+            if (pan->gain.d[i][j] < -128 || pan->gain.d[i][j] > 128)
+                av_log(ctx, AV_LOG_WARNING,
+                    "Gain #%d->#%d too large, clamped\n", j, i);
+            pan->gain.i[i][j] = av_clipf(pan->gain.d[i][j], -128, 128) * 256.0;
+        }
+    }
+    return 0;
+}
+
+
+static void filter_samples(AVFilterLink *inlink, AVFilterBufferRef *insamples)
+{
+    PanContext *const pan = inlink->dst->priv;
+    int i, o, n = insamples->audio->nb_samples;
+
+    /* input */
+    const int16_t *in     = (int16_t *)insamples->data[0];
+    const int16_t *in_end = in + n * pan->nb_input_channels;
+
+    /* output */
+    AVFilterLink *const outlink = inlink->dst->outputs[0];
+    AVFilterBufferRef *outsamples = avfilter_get_audio_buffer(outlink, AV_PERM_WRITE, n);
+    int16_t *out = (int16_t *)outsamples->data[0];
+
+    for (; in < in_end; in += pan->nb_input_channels) {
+        for (o = 0; o < pan->nb_output_channels; o++) {
+            int v = 0;
+            for (i = 0; i < pan->nb_input_channels; i++)
+                v += pan->gain.i[o][i] * in[i];
+            *(out++) = v >> 8;
+        }
+    }
+
+    avfilter_filter_samples(outlink, outsamples);
+    avfilter_unref_buffer(insamples);
+}
+
+AVFilter avfilter_af_pan = {
+    .name          = "pan",
+    .description   = NULL_IF_CONFIG_SMALL("Remix channels with coefficients (panning)"),
+    .priv_size     = sizeof(PanContext),
+    .init          = init,
+    .query_formats = query_formats,
+
+    .inputs    = (const AVFilterPad[]) {
+        { .name             = "default",
+          .type             = AVMEDIA_TYPE_AUDIO,
+          .config_props     = config_props,
+          .filter_samples   = filter_samples,
+          .min_perms        = AV_PERM_READ, },
+        { .name = NULL}
+    },
+    .outputs   = (const AVFilterPad[]) {
+        { .name             = "default",
+          .type             = AVMEDIA_TYPE_AUDIO, },
+        { .name = NULL}
+    },
+};
index e0e5c6f..c4b6972 100644 (file)
@@ -40,6 +40,7 @@ void avfilter_register_all(void)
     REGISTER_FILTER (ARESAMPLE,   aresample,   af);
     REGISTER_FILTER (ASHOWINFO,   ashowinfo,   af);
     REGISTER_FILTER (EARWAX,      earwax,      af);
+    REGISTER_FILTER (PAN,         pan,         af);
     REGISTER_FILTER (VOLUME,      volume,      af);
 
     REGISTER_FILTER (ABUFFER,     abuffer,     asrc);
index 3ce33bb..5e1cfdb 100644 (file)
@@ -29,8 +29,8 @@
 #include "libavutil/rational.h"
 
 #define LIBAVFILTER_VERSION_MAJOR  2
-#define LIBAVFILTER_VERSION_MINOR 48
-#define LIBAVFILTER_VERSION_MICRO  1
+#define LIBAVFILTER_VERSION_MINOR 49
+#define LIBAVFILTER_VERSION_MICRO  0
 
 #define LIBAVFILTER_VERSION_INT AV_VERSION_INT(LIBAVFILTER_VERSION_MAJOR, \
                                                LIBAVFILTER_VERSION_MINOR, \