lavr: add a public function for setting a custom channel map

This allows reordering, duplication, and silencing of input channels.
This commit is contained in:
Justin Ruggles 2012-12-19 14:58:57 -05:00
parent 4d68269d58
commit 074a00d192
11 changed files with 305 additions and 27 deletions

View File

@ -13,6 +13,10 @@ libavutil: 2012-10-22
API changes, most recent first:
2013-xx-xx - xxxxxxx - lavr 1.1.0
Add avresample_set_channel_mapping() for input channel reordering,
duplication, and silencing.
2012-xx-xx - xxxxxxx - lavu 52.2.1 - avstring.h
Add av_basename() and av_dirname().

View File

@ -50,6 +50,7 @@ struct AudioConvert {
DitherContext *dc;
enum AVSampleFormat in_fmt;
enum AVSampleFormat out_fmt;
int apply_map;
int channels;
int planes;
int ptr_align;
@ -259,7 +260,8 @@ void ff_audio_convert_free(AudioConvert **ac)
AudioConvert *ff_audio_convert_alloc(AVAudioResampleContext *avr,
enum AVSampleFormat out_fmt,
enum AVSampleFormat in_fmt,
int channels, int sample_rate)
int channels, int sample_rate,
int apply_map)
{
AudioConvert *ac;
int in_planar, out_planar;
@ -272,11 +274,13 @@ AudioConvert *ff_audio_convert_alloc(AVAudioResampleContext *avr,
ac->out_fmt = out_fmt;
ac->in_fmt = in_fmt;
ac->channels = channels;
ac->apply_map = apply_map;
if (avr->dither_method != AV_RESAMPLE_DITHER_NONE &&
av_get_packed_sample_fmt(out_fmt) == AV_SAMPLE_FMT_S16 &&
av_get_bytes_per_sample(in_fmt) > 2) {
ac->dc = ff_dither_alloc(avr, out_fmt, in_fmt, channels, sample_rate);
ac->dc = ff_dither_alloc(avr, out_fmt, in_fmt, channels, sample_rate,
apply_map);
if (!ac->dc) {
av_free(ac);
return NULL;
@ -309,6 +313,7 @@ int ff_audio_convert(AudioConvert *ac, AudioData *out, AudioData *in)
{
int use_generic = 1;
int len = in->nb_samples;
int p;
if (ac->dc) {
/* dithered conversion */
@ -335,9 +340,46 @@ int ff_audio_convert(AudioConvert *ac, AudioData *out, AudioData *in)
av_get_sample_fmt_name(ac->out_fmt),
use_generic ? ac->func_descr_generic : ac->func_descr);
if (ac->apply_map) {
ChannelMapInfo *map = &ac->avr->ch_map_info;
if (!av_sample_fmt_is_planar(ac->out_fmt)) {
av_log(ac->avr, AV_LOG_ERROR, "cannot remap packed format during conversion\n");
return AVERROR(EINVAL);
}
if (map->do_remap) {
if (av_sample_fmt_is_planar(ac->in_fmt)) {
conv_func_flat *convert = use_generic ? ac->conv_flat_generic :
ac->conv_flat;
for (p = 0; p < ac->planes; p++)
if (map->channel_map[p] >= 0)
convert(out->data[p], in->data[map->channel_map[p]], len);
} else {
uint8_t *data[AVRESAMPLE_MAX_CHANNELS];
conv_func_deinterleave *convert = use_generic ?
ac->conv_deinterleave_generic :
ac->conv_deinterleave;
for (p = 0; p < ac->channels; p++)
data[map->input_map[p]] = out->data[p];
convert(data, in->data[0], len, ac->channels);
}
}
if (map->do_copy || map->do_zero) {
for (p = 0; p < ac->planes; p++) {
if (map->channel_copy[p])
memcpy(out->data[p], out->data[map->channel_copy[p]],
len * out->stride);
else if (map->channel_zero[p])
av_samples_set_silence(&out->data[p], 0, len, 1, ac->out_fmt);
}
}
} else {
switch (ac->func_type) {
case CONV_FUNC_TYPE_FLAT: {
int p;
if (!in->is_planar)
len *= in->channels;
if (use_generic) {
@ -362,6 +404,7 @@ int ff_audio_convert(AudioConvert *ac, AudioData *out, AudioData *in)
ac->conv_deinterleave(out->data, in->data[0], len, ac->channels);
break;
}
}
out->nb_samples = in->nb_samples;
return 0;

View File

@ -58,12 +58,14 @@ void ff_audio_convert_set_func(AudioConvert *ac, enum AVSampleFormat out_fmt,
* @param in_fmt input sample format
* @param channels number of channels
* @param sample_rate sample rate (used for dithering)
* @param apply_map apply channel map during conversion
* @return newly-allocated AudioConvert context
*/
AudioConvert *ff_audio_convert_alloc(AVAudioResampleContext *avr,
enum AVSampleFormat out_fmt,
enum AVSampleFormat in_fmt,
int channels, int sample_rate);
int channels, int sample_rate,
int apply_map);
/**
* Free AudioConvert.

View File

@ -213,7 +213,7 @@ void ff_audio_data_free(AudioData **a)
av_freep(a);
}
int ff_audio_data_copy(AudioData *dst, AudioData *src)
int ff_audio_data_copy(AudioData *dst, AudioData *src, ChannelMapInfo *map)
{
int ret, p;
@ -221,6 +221,11 @@ int ff_audio_data_copy(AudioData *dst, AudioData *src)
if (dst->sample_fmt != src->sample_fmt || dst->channels < src->channels)
return AVERROR(EINVAL);
if (map && !src->is_planar) {
av_log(src, AV_LOG_ERROR, "cannot remap packed format during copy\n");
return AVERROR(EINVAL);
}
/* if the input is empty, just empty the output */
if (!src->nb_samples) {
dst->nb_samples = 0;
@ -233,8 +238,29 @@ int ff_audio_data_copy(AudioData *dst, AudioData *src)
return ret;
/* copy data */
for (p = 0; p < src->planes; p++)
memcpy(dst->data[p], src->data[p], src->nb_samples * src->stride);
if (map) {
if (map->do_remap) {
for (p = 0; p < src->planes; p++) {
if (map->channel_map[p] >= 0)
memcpy(dst->data[p], src->data[map->channel_map[p]],
src->nb_samples * src->stride);
}
}
if (map->do_copy || map->do_zero) {
for (p = 0; p < src->planes; p++) {
if (map->channel_copy[p])
memcpy(dst->data[p], dst->data[map->channel_copy[p]],
src->nb_samples * src->stride);
else if (map->channel_zero[p])
av_samples_set_silence(&dst->data[p], 0, src->nb_samples,
1, dst->sample_fmt);
}
}
} else {
for (p = 0; p < src->planes; p++)
memcpy(dst->data[p], src->data[p], src->nb_samples * src->stride);
}
dst->nb_samples = src->nb_samples;
return 0;

View File

@ -118,9 +118,10 @@ void ff_audio_data_free(AudioData **a);
*
* @param out output AudioData
* @param in input AudioData
* @param map channel map, NULL if not remapping
* @return 0 on success, negative AVERROR value on error
*/
int ff_audio_data_copy(AudioData *out, AudioData *in);
int ff_audio_data_copy(AudioData *out, AudioData *in, ChannelMapInfo *map);
/**
* Append data from one AudioData to the end of another.

View File

@ -258,6 +258,36 @@ int avresample_get_matrix(AVAudioResampleContext *avr, double *matrix,
int avresample_set_matrix(AVAudioResampleContext *avr, const double *matrix,
int stride);
/**
* Set a customized input channel mapping.
*
* This function can only be called when the allocated context is not open.
* Also, the input channel layout must have already been set.
*
* Calling avresample_close() on the context will clear the channel mapping.
*
* The map for each input channel specifies the channel index in the source to
* use for that particular channel, or -1 to mute the channel. Source channels
* can be duplicated by using the same index for multiple input channels.
*
* Examples:
*
* Reordering 5.1 AAC order (C,L,R,Ls,Rs,LFE) to Libav order (L,R,C,LFE,Ls,Rs):
* { 1, 2, 0, 5, 3, 4 }
*
* Muting the 3rd channel in 4-channel input:
* { 0, 1, -1, 3 }
*
* Duplicating the left channel of stereo input:
* { 0, 0 }
*
* @param avr audio resample context
* @param channel_map customized input channel mapping
* @return 0 on success, negative AVERROR code on failure
*/
int avresample_set_channel_mapping(AVAudioResampleContext *avr,
const int *channel_map);
/**
* Set compensation for resampling.
*

View File

@ -53,6 +53,8 @@ typedef struct DitherState {
struct DitherContext {
DitherDSPContext ddsp;
enum AVResampleDitherMethod method;
int apply_map;
ChannelMapInfo *ch_map_info;
int mute_dither_threshold; // threshold for disabling dither
int mute_reset_threshold; // threshold for resetting noise shaping
@ -251,17 +253,23 @@ int ff_convert_dither(DitherContext *c, AudioData *dst, AudioData *src)
return ret;
}
if (src->sample_fmt != AV_SAMPLE_FMT_FLTP) {
if (src->sample_fmt != AV_SAMPLE_FMT_FLTP || c->apply_map) {
/* make sure flt_data is large enough for the input */
ret = ff_audio_data_realloc(c->flt_data, src->nb_samples);
if (ret < 0)
return ret;
flt_data = c->flt_data;
}
if (src->sample_fmt != AV_SAMPLE_FMT_FLTP) {
/* convert input samples to fltp and scale to s16 range */
ret = ff_audio_convert(c->ac_in, flt_data, src);
if (ret < 0)
return ret;
} else if (c->apply_map) {
ret = ff_audio_data_copy(flt_data, src, c->ch_map_info);
if (ret < 0)
return ret;
} else {
flt_data = src;
}
@ -333,7 +341,7 @@ static void dither_init(DitherDSPContext *ddsp,
DitherContext *ff_dither_alloc(AVAudioResampleContext *avr,
enum AVSampleFormat out_fmt,
enum AVSampleFormat in_fmt,
int channels, int sample_rate)
int channels, int sample_rate, int apply_map)
{
AVLFG seed_gen;
DitherContext *c;
@ -350,6 +358,10 @@ DitherContext *ff_dither_alloc(AVAudioResampleContext *avr,
if (!c)
return NULL;
c->apply_map = apply_map;
if (apply_map)
c->ch_map_info = &avr->ch_map_info;
if (avr->dither_method == AV_RESAMPLE_DITHER_TRIANGULAR_NS &&
sample_rate != 48000 && sample_rate != 44100) {
av_log(avr, AV_LOG_WARNING, "sample rate must be 48000 or 44100 Hz "
@ -379,19 +391,20 @@ DitherContext *ff_dither_alloc(AVAudioResampleContext *avr,
goto fail;
c->ac_out = ff_audio_convert_alloc(avr, out_fmt, AV_SAMPLE_FMT_S16P,
channels, sample_rate);
channels, sample_rate, 0);
if (!c->ac_out)
goto fail;
}
if (in_fmt != AV_SAMPLE_FMT_FLTP) {
if (in_fmt != AV_SAMPLE_FMT_FLTP || c->apply_map) {
c->flt_data = ff_audio_data_alloc(channels, 1024, AV_SAMPLE_FMT_FLTP,
"dither flt buffer");
if (!c->flt_data)
goto fail;
}
if (in_fmt != AV_SAMPLE_FMT_FLTP) {
c->ac_in = ff_audio_convert_alloc(avr, AV_SAMPLE_FMT_FLTP, in_fmt,
channels, sample_rate);
channels, sample_rate, c->apply_map);
if (!c->ac_in)
goto fail;
}

View File

@ -66,7 +66,7 @@ typedef struct DitherDSPContext {
DitherContext *ff_dither_alloc(AVAudioResampleContext *avr,
enum AVSampleFormat out_fmt,
enum AVSampleFormat in_fmt,
int channels, int sample_rate);
int channels, int sample_rate, int apply_map);
/**
* Free a DitherContext.

View File

@ -32,6 +32,24 @@ typedef struct AudioConvert AudioConvert;
typedef struct AudioMix AudioMix;
typedef struct ResampleContext ResampleContext;
enum RemapPoint {
REMAP_NONE,
REMAP_IN_COPY,
REMAP_IN_CONVERT,
REMAP_OUT_COPY,
REMAP_OUT_CONVERT,
};
typedef struct ChannelMapInfo {
int channel_map[AVRESAMPLE_MAX_CHANNELS]; /**< source index of each output channel, -1 if not remapped */
int do_remap; /**< remap needed */
int channel_copy[AVRESAMPLE_MAX_CHANNELS]; /**< dest index to copy from */
int do_copy; /**< copy needed */
int channel_zero[AVRESAMPLE_MAX_CHANNELS]; /**< dest index to zero */
int do_zero; /**< zeroing needed */
int input_map[AVRESAMPLE_MAX_CHANNELS]; /**< dest index of each input channel */
} ChannelMapInfo;
struct AVAudioResampleContext {
const AVClass *av_class; /**< AVClass for logging and AVOptions */
@ -65,6 +83,7 @@ struct AVAudioResampleContext {
int resample_needed; /**< resampling is needed */
int in_convert_needed; /**< input sample format conversion is needed */
int out_convert_needed; /**< output sample format conversion is needed */
int in_copy_needed; /**< input data copy is needed */
AudioData *in_buffer; /**< buffer for converted input */
AudioData *resample_out_buffer; /**< buffer for output from resampler */
@ -82,6 +101,10 @@ struct AVAudioResampleContext {
* only used if avresample_set_matrix() is called before avresample_open()
*/
double *mix_matrix;
int use_channel_map;
enum RemapPoint remap_point;
ChannelMapInfo ch_map_info;
};
#endif /* AVRESAMPLE_INTERNAL_H */

View File

@ -96,20 +96,84 @@ int avresample_open(AVAudioResampleContext *avr)
av_get_sample_fmt_name(avr->internal_sample_fmt));
}
/* set sample format conversion parameters */
/* treat all mono as planar for easier comparison */
if (avr->in_channels == 1)
avr->in_sample_fmt = av_get_planar_sample_fmt(avr->in_sample_fmt);
if (avr->out_channels == 1)
avr->out_sample_fmt = av_get_planar_sample_fmt(avr->out_sample_fmt);
avr->in_convert_needed = (avr->resample_needed || avr->mixing_needed) &&
avr->in_sample_fmt != avr->internal_sample_fmt;
/* we may need to add an extra conversion in order to remap channels if
the output format is not planar */
if (avr->use_channel_map && !avr->mixing_needed && !avr->resample_needed &&
!av_sample_fmt_is_planar(avr->out_sample_fmt)) {
avr->internal_sample_fmt = av_get_planar_sample_fmt(avr->out_sample_fmt);
}
/* set sample format conversion parameters */
if (avr->resample_needed || avr->mixing_needed)
avr->in_convert_needed = avr->in_sample_fmt != avr->internal_sample_fmt;
else
avr->in_convert_needed = avr->use_channel_map &&
!av_sample_fmt_is_planar(avr->out_sample_fmt);
if (avr->resample_needed || avr->mixing_needed || avr->in_convert_needed)
avr->out_convert_needed = avr->internal_sample_fmt != avr->out_sample_fmt;
else
avr->out_convert_needed = avr->in_sample_fmt != avr->out_sample_fmt;
avr->in_copy_needed = !avr->in_convert_needed && (avr->mixing_needed ||
(avr->use_channel_map && avr->resample_needed));
if (avr->use_channel_map) {
if (avr->in_copy_needed) {
avr->remap_point = REMAP_IN_COPY;
av_dlog(avr, "remap channels during in_copy\n");
} else if (avr->in_convert_needed) {
avr->remap_point = REMAP_IN_CONVERT;
av_dlog(avr, "remap channels during in_convert\n");
} else if (avr->out_convert_needed) {
avr->remap_point = REMAP_OUT_CONVERT;
av_dlog(avr, "remap channels during out_convert\n");
} else {
avr->remap_point = REMAP_OUT_COPY;
av_dlog(avr, "remap channels during out_copy\n");
}
#ifdef DEBUG
{
int ch;
av_dlog(avr, "output map: ");
if (avr->ch_map_info.do_remap)
for (ch = 0; ch < avr->in_channels; ch++)
av_dlog(avr, " % 2d", avr->ch_map_info.channel_map[ch]);
else
av_dlog(avr, "n/a");
av_dlog(avr, "\n");
av_dlog(avr, "copy map: ");
if (avr->ch_map_info.do_copy)
for (ch = 0; ch < avr->in_channels; ch++)
av_dlog(avr, " % 2d", avr->ch_map_info.channel_copy[ch]);
else
av_dlog(avr, "n/a");
av_dlog(avr, "\n");
av_dlog(avr, "zero map: ");
if (avr->ch_map_info.do_zero)
for (ch = 0; ch < avr->in_channels; ch++)
av_dlog(avr, " % 2d", avr->ch_map_info.channel_zero[ch]);
else
av_dlog(avr, "n/a");
av_dlog(avr, "\n");
av_dlog(avr, "input map: ");
for (ch = 0; ch < avr->in_channels; ch++)
av_dlog(avr, " % 2d", avr->ch_map_info.input_map[ch]);
av_dlog(avr, "\n");
}
#endif
} else
avr->remap_point = REMAP_NONE;
/* allocate buffers */
if (avr->mixing_needed || avr->in_convert_needed) {
if (avr->in_copy_needed || avr->in_convert_needed) {
avr->in_buffer = ff_audio_data_alloc(FFMAX(avr->in_channels, avr->out_channels),
0, avr->internal_sample_fmt,
"in_buffer");
@ -146,7 +210,8 @@ int avresample_open(AVAudioResampleContext *avr)
if (avr->in_convert_needed) {
avr->ac_in = ff_audio_convert_alloc(avr, avr->internal_sample_fmt,
avr->in_sample_fmt, avr->in_channels,
avr->in_sample_rate);
avr->in_sample_rate,
avr->remap_point == REMAP_IN_CONVERT);
if (!avr->ac_in) {
ret = AVERROR(ENOMEM);
goto error;
@ -160,7 +225,8 @@ int avresample_open(AVAudioResampleContext *avr)
src_fmt = avr->in_sample_fmt;
avr->ac_out = ff_audio_convert_alloc(avr, avr->out_sample_fmt, src_fmt,
avr->out_channels,
avr->out_sample_rate);
avr->out_sample_rate,
avr->remap_point == REMAP_OUT_CONVERT);
if (!avr->ac_out) {
ret = AVERROR(ENOMEM);
goto error;
@ -200,6 +266,8 @@ void avresample_close(AVAudioResampleContext *avr)
ff_audio_resample_free(&avr->resample);
ff_audio_mix_free(&avr->am);
av_freep(&avr->mix_matrix);
avr->use_channel_map = 0;
}
void avresample_free(AVAudioResampleContext **avr)
@ -242,7 +310,9 @@ static int handle_buffered_output(AVAudioResampleContext *avr,
data in the output FIFO */
av_dlog(avr, "[copy] %s to output\n", converted->name);
output->nb_samples = 0;
ret = ff_audio_data_copy(output, converted);
ret = ff_audio_data_copy(output, converted,
avr->remap_point == REMAP_OUT_COPY ?
&avr->ch_map_info : NULL);
if (ret < 0)
return ret;
av_dlog(avr, "[end conversion]\n");
@ -306,11 +376,24 @@ int attribute_align_arg avresample_convert(AVAudioResampleContext *avr,
/* in some rare cases we can copy input to output and upmix
directly in the output buffer */
av_dlog(avr, "[copy] %s to output\n", current_buffer->name);
ret = ff_audio_data_copy(&output_buffer, current_buffer);
ret = ff_audio_data_copy(&output_buffer, current_buffer,
avr->remap_point == REMAP_OUT_COPY ?
&avr->ch_map_info : NULL);
if (ret < 0)
return ret;
current_buffer = &output_buffer;
} else if (avr->mixing_needed || avr->in_convert_needed) {
} else if (avr->remap_point == REMAP_OUT_COPY &&
(!direct_output || out_samples < in_samples)) {
/* if remapping channels during output copy, we may need to
* use an intermediate buffer in order to remap before adding
* samples to the output fifo */
av_dlog(avr, "[copy] %s to out_buffer\n", current_buffer->name);
ret = ff_audio_data_copy(avr->out_buffer, current_buffer,
&avr->ch_map_info);
if (ret < 0)
return ret;
current_buffer = avr->out_buffer;
} else if (avr->in_copy_needed || avr->in_convert_needed) {
/* if needed, copy or convert input to in_buffer, and downmix if
applicable */
if (avr->in_convert_needed) {
@ -325,7 +408,9 @@ int attribute_align_arg avresample_convert(AVAudioResampleContext *avr,
return ret;
} else {
av_dlog(avr, "[copy] %s to in_buffer\n", current_buffer->name);
ret = ff_audio_data_copy(avr->in_buffer, current_buffer);
ret = ff_audio_data_copy(avr->in_buffer, current_buffer,
avr->remap_point == REMAP_IN_COPY ?
&avr->ch_map_info : NULL);
if (ret < 0)
return ret;
}
@ -470,6 +555,57 @@ int avresample_set_matrix(AVAudioResampleContext *avr, const double *matrix,
return 0;
}
int avresample_set_channel_mapping(AVAudioResampleContext *avr,
const int *channel_map)
{
ChannelMapInfo *info = &avr->ch_map_info;
int in_channels, ch, i;
in_channels = av_get_channel_layout_nb_channels(avr->in_channel_layout);
if (in_channels <= 0 || in_channels > AVRESAMPLE_MAX_CHANNELS) {
av_log(avr, AV_LOG_ERROR, "Invalid input channel layout\n");
return AVERROR(EINVAL);
}
memset(info, 0, sizeof(*info));
memset(info->input_map, -1, sizeof(info->input_map));
for (ch = 0; ch < in_channels; ch++) {
if (channel_map[ch] >= in_channels) {
av_log(avr, AV_LOG_ERROR, "Invalid channel map\n");
return AVERROR(EINVAL);
}
if (channel_map[ch] < 0) {
info->channel_zero[ch] = 1;
info->channel_map[ch] = -1;
info->do_zero = 1;
} else if (info->input_map[channel_map[ch]] >= 0) {
info->channel_copy[ch] = info->input_map[channel_map[ch]];
info->channel_map[ch] = -1;
info->do_copy = 1;
} else {
info->channel_map[ch] = channel_map[ch];
info->input_map[channel_map[ch]] = ch;
info->do_remap = 1;
}
}
/* Fill-in unmapped input channels with unmapped output channels.
This is used when remapping during conversion from interleaved to
planar format. */
for (ch = 0, i = 0; ch < in_channels && i < in_channels; ch++, i++) {
while (ch < in_channels && info->input_map[ch] >= 0)
ch++;
while (i < in_channels && info->channel_map[i] >= 0)
i++;
if (ch >= in_channels || i >= in_channels)
break;
info->input_map[ch] = i;
}
avr->use_channel_map = 1;
return 0;
}
int avresample_available(AVAudioResampleContext *avr)
{
return av_audio_fifo_size(avr->out_fifo);

View File

@ -20,8 +20,8 @@
#define AVRESAMPLE_VERSION_H
#define LIBAVRESAMPLE_VERSION_MAJOR 1
#define LIBAVRESAMPLE_VERSION_MINOR 0
#define LIBAVRESAMPLE_VERSION_MICRO 1
#define LIBAVRESAMPLE_VERSION_MINOR 1
#define LIBAVRESAMPLE_VERSION_MICRO 0
#define LIBAVRESAMPLE_VERSION_INT AV_VERSION_INT(LIBAVRESAMPLE_VERSION_MAJOR, \
LIBAVRESAMPLE_VERSION_MINOR, \