lavr: add option for dithering during sample format conversion to s16

libavresample/Makefile

@@ -8,6 +8,7 @@ OBJS = audio_convert.o                                                  \
        audio_data.o                                                     \
        audio_mix.o                                                      \
        audio_mix_matrix.o                                               \
+       dither.o                                                         \
        options.o                                                        \
        resample.o                                                       \
        utils.o                                                          \

libavresample/audio_convert.c

@@ -29,6 +29,8 @@
 #include "libavutil/samplefmt.h"
 #include "audio_convert.h"
 #include "audio_data.h"
+#include "dither.h"
+#include "internal.h"
 
 enum ConvFuncType {
     CONV_FUNC_TYPE_FLAT,
@@ -46,6 +48,7 @@ typedef void (conv_func_deinterleave)(uint8_t **out, const uint8_t *in, int len,
 
 struct AudioConvert {
     AVAudioResampleContext *avr;
+    DitherContext *dc;
     enum AVSampleFormat in_fmt;
     enum AVSampleFormat out_fmt;
     int channels;
@@ -246,10 +249,18 @@ static void set_generic_function(AudioConvert *ac)
     SET_CONV_FUNC_GROUP(AV_SAMPLE_FMT_DBL, AV_SAMPLE_FMT_DBL)
 }
 
+void ff_audio_convert_free(AudioConvert **ac)
+{
+    if (!*ac)
+        return;
+    ff_dither_free(&(*ac)->dc);
+    av_freep(ac);
+}
+
 AudioConvert *ff_audio_convert_alloc(AVAudioResampleContext *avr,
                                      enum AVSampleFormat out_fmt,
                                      enum AVSampleFormat in_fmt,
-                                     int channels)
+                                     int channels, int sample_rate)
 {
     AudioConvert *ac;
     int in_planar, out_planar;
@@ -263,6 +274,17 @@ AudioConvert *ff_audio_convert_alloc(AVAudioResampleContext *avr,
     ac->in_fmt   = in_fmt;
     ac->channels = channels;
 
+    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);
+        if (!ac->dc) {
+            av_free(ac);
+            return NULL;
+        }
+        return ac;
+    }
+
     in_planar  = av_sample_fmt_is_planar(in_fmt);
     out_planar = av_sample_fmt_is_planar(out_fmt);
 
@@ -289,6 +311,15 @@ int ff_audio_convert(AudioConvert *ac, AudioData *out, AudioData *in)
     int use_generic = 1;
     int len         = in->nb_samples;
 
+    if (ac->dc) {
+        /* dithered conversion */
+        av_dlog(ac->avr, "%d samples - audio_convert: %s to %s (dithered)\n",
+                len, av_get_sample_fmt_name(ac->in_fmt),
+                av_get_sample_fmt_name(ac->out_fmt));
+
+        return ff_convert_dither(ac->dc, out, in);
+    }
+
     /* determine whether to use the optimized function based on pointer and
        samples alignment in both the input and output */
     if (ac->has_optimized_func) {

libavresample/audio_convert.h

@@ -54,16 +54,26 @@ void ff_audio_convert_set_func(AudioConvert *ac, enum AVSampleFormat out_fmt,
 /**
  * Allocate and initialize AudioConvert context for sample format conversion.
  *
- * @param avr      AVAudioResampleContext
- * @param out_fmt  output sample format
- * @param in_fmt   input sample format
- * @param channels number of channels
- * @return         newly-allocated AudioConvert context
+ * @param avr         AVAudioResampleContext
+ * @param out_fmt     output sample format
+ * @param in_fmt      input sample format
+ * @param channels    number of channels
+ * @param sample_rate sample rate (used for dithering)
+ * @return            newly-allocated AudioConvert context
  */
 AudioConvert *ff_audio_convert_alloc(AVAudioResampleContext *avr,
                                      enum AVSampleFormat out_fmt,
                                      enum AVSampleFormat in_fmt,
-                                     int channels);
+                                     int channels, int sample_rate);
+
+/**
+ * Free AudioConvert.
+ *
+ * The AudioConvert must have been previously allocated with ff_audio_convert_alloc().
+ *
+ * @param ac  AudioConvert struct
+ */
+void ff_audio_convert_free(AudioConvert **ac);
 
 /**
  * Convert audio data from one sample format to another.

libavresample/avresample.h

@@ -119,6 +119,15 @@ enum AVResampleFilterType {
     AV_RESAMPLE_FILTER_TYPE_KAISER,             /**< Kaiser Windowed Sinc */
 };
 
+enum AVResampleDitherMethod {
+    AV_RESAMPLE_DITHER_NONE,            /**< Do not use dithering */
+    AV_RESAMPLE_DITHER_RECTANGULAR,     /**< Rectangular Dither */
+    AV_RESAMPLE_DITHER_TRIANGULAR,      /**< Triangular Dither*/
+    AV_RESAMPLE_DITHER_TRIANGULAR_HP,   /**< Triangular Dither with High Pass */
+    AV_RESAMPLE_DITHER_TRIANGULAR_NS,   /**< Triangular Dither with Noise Shaping */
+    AV_RESAMPLE_DITHER_NB,              /**< Number of dither types. Not part of ABI. */
+};
+
 /**
  * Return the LIBAVRESAMPLE_VERSION_INT constant.
  */

libavresample/dither.h

+/*
+ * Copyright (c) 2012 Justin Ruggles <justin.ruggles@gmail.com>
+ *
+ * This file is part of Libav.
+ *
+ * Libav 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.
+ *
+ * Libav 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 Libav; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+#ifndef AVRESAMPLE_DITHER_H
+#define AVRESAMPLE_DITHER_H
+
+#include "avresample.h"
+#include "audio_data.h"
+
+typedef struct DitherContext DitherContext;
+
+typedef struct DitherDSPContext {
+    /**
+     * Convert samples from flt to s16 with added dither noise.
+     *
+     * @param dst    destination float array, range -0.5 to 0.5
+     * @param src    source int array, range INT_MIN to INT_MAX.
+     * @param dither float dither noise array
+     * @param len    number of samples
+     */
+    void (*quantize)(int16_t *dst, const float *src, float *dither, int len);
+
+    int ptr_align;      ///< src and dst constraits for quantize()
+    int samples_align;  ///< len constraits for quantize()
+
+    /**
+     * Convert dither noise from int to float with triangular distribution.
+     *
+     * @param dst  destination float array, range -0.5 to 0.5
+     *             constraints: 32-byte aligned
+     * @param src0 source int array, range INT_MIN to INT_MAX.
+     *             the array size is len * 2
+     *             constraints: 32-byte aligned
+     * @param len  number of output noise samples
+     *             constraints: multiple of 16
+     */
+    void (*dither_int_to_float)(float *dst, int *src0, int len);
+} DitherDSPContext;
+
+/**
+ * Allocate and initialize a DitherContext.
+ *
+ * The parameters in the AVAudioResampleContext are used to initialize the
+ * DitherContext.
+ *
+ * @param avr  AVAudioResampleContext
+ * @return     newly-allocated DitherContext
+ */
+DitherContext *ff_dither_alloc(AVAudioResampleContext *avr,
+                               enum AVSampleFormat out_fmt,
+                               enum AVSampleFormat in_fmt,
+                               int channels, int sample_rate);
+
+/**
+ * Free a DitherContext.
+ *
+ * @param c  DitherContext
+ */
+void ff_dither_free(DitherContext **c);
+
+/**
+ * Convert audio sample format with dithering.
+ *
+ * @param c    DitherContext
+ * @param dst  destination audio data
+ * @param src  source audio data
+ * @return     0 if ok, negative AVERROR code on failure
+ */
+int ff_convert_dither(DitherContext *c, AudioData *dst, AudioData *src);
+
+#endif /* AVRESAMPLE_DITHER_H */

libavresample/internal.h

@@ -53,6 +53,7 @@ struct AVAudioResampleContext {
     double cutoff;                              /**< resampling cutoff frequency. 1.0 corresponds to half the output sample rate */
     enum AVResampleFilterType filter_type;      /**< resampling filter type */
     int kaiser_beta;                            /**< beta value for Kaiser window (only applicable if filter_type == AV_FILTER_TYPE_KAISER) */
+    enum AVResampleDitherMethod dither_method;  /**< dither method          */
 
     int in_channels;        /**< number of input channels                   */
     int out_channels;       /**< number of output channels                  */

libavresample/options.c

@@ -63,6 +63,12 @@ static const AVOption options[] = {
         { "blackman_nuttall", "Blackman Nuttall Windowed Sinc", 0, AV_OPT_TYPE_CONST, { .i64 = AV_RESAMPLE_FILTER_TYPE_BLACKMAN_NUTTALL }, INT_MIN, INT_MAX, PARAM, "filter_type" },
         { "kaiser",           "Kaiser Windowed Sinc",           0, AV_OPT_TYPE_CONST, { .i64 = AV_RESAMPLE_FILTER_TYPE_KAISER           }, INT_MIN, INT_MAX, PARAM, "filter_type" },
     { "kaiser_beta",            "Kaiser Window Beta",       OFFSET(kaiser_beta),            AV_OPT_TYPE_INT,    { .i64 = 9              }, 2,                    16,                     PARAM },
+    { "dither_method",          "Dither Method",            OFFSET(dither_method),          AV_OPT_TYPE_INT,    { .i64 = AV_RESAMPLE_DITHER_NONE }, 0, AV_RESAMPLE_DITHER_NB-1, PARAM, "dither_method"},
+        {"none",          "No Dithering",                         0, AV_OPT_TYPE_CONST, { .i64 = AV_RESAMPLE_DITHER_NONE          }, INT_MIN, INT_MAX, PARAM, "dither_method"},
+        {"rectangular",   "Rectangular Dither",                   0, AV_OPT_TYPE_CONST, { .i64 = AV_RESAMPLE_DITHER_RECTANGULAR   }, INT_MIN, INT_MAX, PARAM, "dither_method"},
+        {"triangular",    "Triangular Dither",                    0, AV_OPT_TYPE_CONST, { .i64 = AV_RESAMPLE_DITHER_TRIANGULAR    }, INT_MIN, INT_MAX, PARAM, "dither_method"},
+        {"triangular_hp", "Triangular Dither With High Pass",     0, AV_OPT_TYPE_CONST, { .i64 = AV_RESAMPLE_DITHER_TRIANGULAR_HP }, INT_MIN, INT_MAX, PARAM, "dither_method"},
+        {"triangular_ns", "Triangular Dither With Noise Shaping", 0, AV_OPT_TYPE_CONST, { .i64 = AV_RESAMPLE_DITHER_TRIANGULAR_NS }, INT_MIN, INT_MAX, PARAM, "dither_method"},
     { NULL },
 };
 

libavresample/utils.c

@@ -142,7 +142,8 @@ int avresample_open(AVAudioResampleContext *avr)
     /* setup contexts */
     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_fmt, avr->in_channels,
+                                            avr->in_sample_rate);
         if (!avr->ac_in) {
             ret = AVERROR(ENOMEM);
             goto error;
@@ -155,7 +156,8 @@ int avresample_open(AVAudioResampleContext *avr)
         else
             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_channels,
+                                             avr->out_sample_rate);
         if (!avr->ac_out) {
             ret = AVERROR(ENOMEM);
             goto error;
@@ -190,8 +192,8 @@ void avresample_close(AVAudioResampleContext *avr)
     ff_audio_data_free(&avr->out_buffer);
     av_audio_fifo_free(avr->out_fifo);
     avr->out_fifo = NULL;
-    av_freep(&avr->ac_in);
-    av_freep(&avr->ac_out);
+    ff_audio_convert_free(&avr->ac_in);
+    ff_audio_convert_free(&avr->ac_out);
     ff_audio_resample_free(&avr->resample);
     ff_audio_mix_free(&avr->am);
     av_freep(&avr->mix_matrix);

libavresample/version.h

@@ -21,7 +21,7 @@
 
 #define LIBAVRESAMPLE_VERSION_MAJOR  1
 #define LIBAVRESAMPLE_VERSION_MINOR  0
-#define LIBAVRESAMPLE_VERSION_MICRO  0
+#define LIBAVRESAMPLE_VERSION_MICRO  1
 
 #define LIBAVRESAMPLE_VERSION_INT  AV_VERSION_INT(LIBAVRESAMPLE_VERSION_MAJOR, \
                                                   LIBAVRESAMPLE_VERSION_MINOR, \