... | ... |
@@ -1544,6 +1544,164 @@ Optional. It should have a value much less than 1 (e.g. 0.05 or 0.02) and is |
1544 | 1544 |
used to prevent clipping. |
1545 | 1545 |
@end table |
1546 | 1546 |
|
1547 |
+@section dynaudnorm |
|
1548 |
+Dynamic Audio Normalizer. |
|
1549 |
+ |
|
1550 |
+This filter applies a certain amount of gain to the input audio in order |
|
1551 |
+to bring its peak magnitude to a target level (e.g. 0 dBFS). However, in |
|
1552 |
+contrast to more "simple" normalization algorithms, the Dynamic Audio |
|
1553 |
+Normalizer *dynamically* re-adjusts the gain factor to the input audio. |
|
1554 |
+This allows for applying extra gain to the "quiet" sections of the audio |
|
1555 |
+while avoiding distortions or clipping the "loud" sections. In other words: |
|
1556 |
+The Dynamic Audio Normalizer will "even out" the volume of quiet and loud |
|
1557 |
+sections, in the sense that the volume of each section is brought to the |
|
1558 |
+same target level. Note, however, that the Dynamic Audio Normalizer achieves |
|
1559 |
+this goal *without* applying "dynamic range compressing". It will retain 100% |
|
1560 |
+of the dynamic range *within* each section of the audio file. |
|
1561 |
+ |
|
1562 |
+@table @option |
|
1563 |
+@item f |
|
1564 |
+Set the frame length in milliseconds. In range from 10 to 8000 milliseconds. |
|
1565 |
+Default is 500 milliseconds. |
|
1566 |
+The Dynamic Audio Normalizer processes the input audio in small chunks, |
|
1567 |
+referred to as frames. This is required, because a peak magnitude has no |
|
1568 |
+meaning for just a single sample value. Instead, we need to determine the |
|
1569 |
+peak magnitude for a contiguous sequence of sample values. While a "standard" |
|
1570 |
+normalizer would simply use the peak magnitude of the complete file, the |
|
1571 |
+Dynamic Audio Normalizer determines the peak magnitude individually for each |
|
1572 |
+frame. The length of a frame is specified in milliseconds. By default, the |
|
1573 |
+Dynamic Audio Normalizer uses a frame length of 500 milliseconds, which has |
|
1574 |
+been found to give good results with most files. |
|
1575 |
+Note that the exact frame length, in number of samples, will be determined |
|
1576 |
+automatically, based on the sampling rate of the individual input audio file. |
|
1577 |
+ |
|
1578 |
+@item g |
|
1579 |
+Set the Gaussian filter window size. In range from 3 to 301, must be odd |
|
1580 |
+number. Default is 31. |
|
1581 |
+Probably the most important parameter of the Dynamic Audio Normalizer is the |
|
1582 |
+@code{window size} of the Gaussian smoothing filter. The filter's window size |
|
1583 |
+is specified in frames, centered around the current frame. For the sake of |
|
1584 |
+simplicity, this must be an odd number. Consequently, the default value of 31 |
|
1585 |
+takes into account the current frame, as well as the 15 preceding frames and |
|
1586 |
+the 15 subsequent frames. Using a larger window results in a stronger |
|
1587 |
+smoothing effect and thus in less gain variation, i.e. slower gain |
|
1588 |
+adaptation. Conversely, using a smaller window results in a weaker smoothing |
|
1589 |
+effect and thus in more gain variation, i.e. faster gain adaptation. |
|
1590 |
+In other words, the more you increase this value, the more the Dynamic Audio |
|
1591 |
+Normalizer will behave like a "traditional" normalization filter. On the |
|
1592 |
+contrary, the more you decrease this value, the more the Dynamic Audio |
|
1593 |
+Normalizer will behave like a dynamic range compressor. |
|
1594 |
+ |
|
1595 |
+@item p |
|
1596 |
+Set the target peak value. This specifies the highest permissible magnitude |
|
1597 |
+level for the normalized audio input. This filter will try to approach the |
|
1598 |
+target peak magnitude as closely as possible, but at the same time it also |
|
1599 |
+makes sure that the normalized signal will never exceed the peak magnitude. |
|
1600 |
+A frame's maximum local gain factor is imposed directly by the target peak |
|
1601 |
+magnitude. The default value is 0.95 and thus leaves a headroom of 5%*. |
|
1602 |
+It is not recommended to go above this value. |
|
1603 |
+ |
|
1604 |
+@item m |
|
1605 |
+Set the maximum gain factor. In range from 1.0 to 100.0. Default is 10.0. |
|
1606 |
+The Dynamic Audio Normalizer determines the maximum possible (local) gain |
|
1607 |
+factor for each input frame, i.e. the maximum gain factor that does not |
|
1608 |
+result in clipping or distortion. The maximum gain factor is determined by |
|
1609 |
+the frame's highest magnitude sample. However, the Dynamic Audio Normalizer |
|
1610 |
+additionally bounds the frame's maximum gain factor by a predetermined |
|
1611 |
+(global) maximum gain factor. This is done in order to avoid excessive gain |
|
1612 |
+factors in "silent" or almost silent frames. By default, the maximum gain |
|
1613 |
+factor is 10.0, For most inputs the default value should be sufficient and |
|
1614 |
+it usually is not recommended to increase this value. Though, for input |
|
1615 |
+with an extremely low overall volume level, it may be necessary to allow even |
|
1616 |
+higher gain factors. Note, however, that the Dynamic Audio Normalizer does |
|
1617 |
+not simply apply a "hard" threshold (i.e. cut off values above the threshold). |
|
1618 |
+Instead, a "sigmoid" threshold function will be applied. This way, the |
|
1619 |
+gain factors will smoothly approach the threshold value, but never exceed that |
|
1620 |
+value. |
|
1621 |
+ |
|
1622 |
+@item r |
|
1623 |
+Set the target RMS. In range from 0.0 to 1.0. Default is 0.0 - disabled. |
|
1624 |
+By default, the Dynamic Audio Normalizer performs "peak" normalization. |
|
1625 |
+This means that the maximum local gain factor for each frame is defined |
|
1626 |
+(only) by the frame's highest magnitude sample. This way, the samples can |
|
1627 |
+be amplified as much as possible without exceeding the maximum signal |
|
1628 |
+level, i.e. without clipping. Optionally, however, the Dynamic Audio |
|
1629 |
+Normalizer can also take into account the frame's root mean square, |
|
1630 |
+abbreviated RMS. In electrical engineering, the RMS is commonly used to |
|
1631 |
+determine the power of a time-varying signal. It is therefore considered |
|
1632 |
+that the RMS is a better approximation of the "perceived loudness" than |
|
1633 |
+just looking at the signal's peak magnitude. Consequently, by adjusting all |
|
1634 |
+frames to a constant RMS value, a uniform "perceived loudness" can be |
|
1635 |
+established. If a target RMS value has been specified, a frame's local gain |
|
1636 |
+factor is defined as the factor that would result in exactly that RMS value. |
|
1637 |
+Note, however, that the maximum local gain factor is still restricted by the |
|
1638 |
+frame's highest magnitude sample, in order to prevent clipping. |
|
1639 |
+ |
|
1640 |
+@item n |
|
1641 |
+Enable channels coupling. By default is enabled. |
|
1642 |
+By default, the Dynamic Audio Normalizer will amplify all channels by the same |
|
1643 |
+amount. This means the same gain factor will be applied to all channels, i.e. |
|
1644 |
+the maximum possible gain factor is determined by the "loudest" channel. |
|
1645 |
+However, in some recordings, it may happen that the volume of the different |
|
1646 |
+channels is uneven, e.g. one channel may be "quieter" than the other one(s). |
|
1647 |
+In this case, this option can be used to disable the channel coupling. This way, |
|
1648 |
+the gain factor will be determined independently for each channel, depending |
|
1649 |
+only on the individual channel's highest magnitude sample. This allows for |
|
1650 |
+harmonizing the volume of the different channels. |
|
1651 |
+ |
|
1652 |
+@item c |
|
1653 |
+Enable DC bias correction. By default is disabled. |
|
1654 |
+An audio signal (in the time domain) is a sequence of sample values. |
|
1655 |
+In the Dynamic Audio Normalizer these sample values are represented in the |
|
1656 |
+-1.0 to 1.0 range, regardless of the original input format. Normally, the |
|
1657 |
+audio signal, or "waveform", should be centered around the zero point. |
|
1658 |
+That means if we calculate the mean value of all samples in a file, or in a |
|
1659 |
+single frame, then the result should be 0.0 or at least very close to that |
|
1660 |
+value. If, however, there is a significant deviation of the mean value from |
|
1661 |
+0.0, in either positive or negative direction, this is referred to as a |
|
1662 |
+DC bias or DC offset. Since a DC bias is clearly undesirable, the Dynamic |
|
1663 |
+Audio Normalizer provides optional DC bias correction. |
|
1664 |
+With DC bias correction enabled, the Dynamic Audio Normalizer will determine |
|
1665 |
+the mean value, or "DC correction" offset, of each input frame and subtract |
|
1666 |
+that value from all of the frame's sample values which ensures those samples |
|
1667 |
+are centered around 0.0 again. Also, in order to avoid "gaps" at the frame |
|
1668 |
+boundaries, the DC correction offset values will be interpolated smoothly |
|
1669 |
+between neighbouring frames. |
|
1670 |
+ |
|
1671 |
+@item b |
|
1672 |
+Enable alternative boundary mode. By default is disabled. |
|
1673 |
+The Dynamic Audio Normalizer takes into account a certain neighbourhood |
|
1674 |
+around each frame. This includes the preceding frames as well as the |
|
1675 |
+subsequent frames. However, for the "boundary" frames, located at the very |
|
1676 |
+beginning and at the very end of the audio file, not all neighbouring |
|
1677 |
+frames are available. In particular, for the first few frames in the audio |
|
1678 |
+file, the preceding frames are not known. And, similarly, for the last few |
|
1679 |
+frames in the audio file, the subsequent frames are not known. Thus, the |
|
1680 |
+question arises which gain factors should be assumed for the missing frames |
|
1681 |
+in the "boundary" region. The Dynamic Audio Normalizer implements two modes |
|
1682 |
+to deal with this situation. The default boundary mode assumes a gain factor |
|
1683 |
+of exactly 1.0 for the missing frames, resulting in a smooth "fade in" and |
|
1684 |
+"fade out" at the beginning and at the end of the input, respectively. |
|
1685 |
+ |
|
1686 |
+@item s |
|
1687 |
+Set the compress factor. In range from 0.0 to 30.0. Default is 0.0. |
|
1688 |
+By default, the Dynamic Audio Normalizer does not apply "traditional" |
|
1689 |
+compression. This means that signal peaks will not be pruned and thus the |
|
1690 |
+full dynamic range will be retained within each local neighbourhood. However, |
|
1691 |
+in some cases it may be desirable to combine the Dynamic Audio Normalizer's |
|
1692 |
+normalization algorithm with a more "traditional" compression. |
|
1693 |
+For this purpose, the Dynamic Audio Normalizer provides an optional compression |
|
1694 |
+(thresholding) function. If (and only if) the compression feature is enabled, |
|
1695 |
+all input frames will be processed by a soft knee thresholding function prior |
|
1696 |
+to the actual normalization process. Put simply, the thresholding function is |
|
1697 |
+going to prune all samples whose magnitude exceeds a certain threshold value. |
|
1698 |
+However, the Dynamic Audio Normalizer does not simply apply a fixed threshold |
|
1699 |
+value. Instead, the threshold value will be adjusted for each individual |
|
1700 |
+frame. |
|
1701 |
+In general, smaller parameters result in stronger compression, and vice versa. |
|
1702 |
+Values below 3.0 are not recommended, because audible distortion may appear. |
|
1703 |
+@end table |
|
1704 |
+ |
|
1547 | 1705 |
@section earwax |
1548 | 1706 |
|
1549 | 1707 |
Make audio easier to listen to on headphones. |
... | ... |
@@ -67,6 +67,7 @@ OBJS-$(CONFIG_CHANNELSPLIT_FILTER) += af_channelsplit.o |
67 | 67 |
OBJS-$(CONFIG_CHORUS_FILTER) += af_chorus.o generate_wave_table.o |
68 | 68 |
OBJS-$(CONFIG_COMPAND_FILTER) += af_compand.o |
69 | 69 |
OBJS-$(CONFIG_DCSHIFT_FILTER) += af_dcshift.o |
70 |
+OBJS-$(CONFIG_DYNAUDNORM_FILTER) += af_dynaudnorm.o |
|
70 | 71 |
OBJS-$(CONFIG_EARWAX_FILTER) += af_earwax.o |
71 | 72 |
OBJS-$(CONFIG_EBUR128_FILTER) += f_ebur128.o |
72 | 73 |
OBJS-$(CONFIG_EQUALIZER_FILTER) += af_biquads.o |
73 | 74 |
new file mode 100644 |
... | ... |
@@ -0,0 +1,734 @@ |
0 |
+/* |
|
1 |
+ * Dynamic Audio Normalizer |
|
2 |
+ * Copyright (c) 2015 LoRd_MuldeR <mulder2@gmx.de>. Some rights reserved. |
|
3 |
+ * |
|
4 |
+ * This file is part of FFmpeg. |
|
5 |
+ * |
|
6 |
+ * FFmpeg is free software; you can redistribute it and/or |
|
7 |
+ * modify it under the terms of the GNU Lesser General Public |
|
8 |
+ * License as published by the Free Software Foundation; either |
|
9 |
+ * version 2.1 of the License, or (at your option) any later version. |
|
10 |
+ * |
|
11 |
+ * FFmpeg is distributed in the hope that it will be useful, |
|
12 |
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of |
|
13 |
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
|
14 |
+ * Lesser General Public License for more details. |
|
15 |
+ * |
|
16 |
+ * You should have received a copy of the GNU Lesser General Public |
|
17 |
+ * License along with FFmpeg; if not, write to the Free Software |
|
18 |
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
|
19 |
+ */ |
|
20 |
+ |
|
21 |
+/** |
|
22 |
+ * @file |
|
23 |
+ * Dynamic Audio Normalizer |
|
24 |
+ */ |
|
25 |
+ |
|
26 |
+#include <float.h> |
|
27 |
+ |
|
28 |
+#include "libavutil/avassert.h" |
|
29 |
+#include "libavutil/opt.h" |
|
30 |
+ |
|
31 |
+#define FF_BUFQUEUE_SIZE 302 |
|
32 |
+#include "libavfilter/bufferqueue.h" |
|
33 |
+ |
|
34 |
+#include "audio.h" |
|
35 |
+#include "avfilter.h" |
|
36 |
+#include "internal.h" |
|
37 |
+ |
|
38 |
+typedef struct cqueue { |
|
39 |
+ double *elements; |
|
40 |
+ int size; |
|
41 |
+ int nb_elements; |
|
42 |
+ int first; |
|
43 |
+} cqueue; |
|
44 |
+ |
|
45 |
+typedef struct DynamicAudioNormalizerContext { |
|
46 |
+ const AVClass *class; |
|
47 |
+ |
|
48 |
+ struct FFBufQueue queue; |
|
49 |
+ |
|
50 |
+ int frame_len; |
|
51 |
+ int frame_len_msec; |
|
52 |
+ int filter_size; |
|
53 |
+ int dc_correction; |
|
54 |
+ int channels_coupled; |
|
55 |
+ int alt_boundary_mode; |
|
56 |
+ |
|
57 |
+ double peak_value; |
|
58 |
+ double max_amplification; |
|
59 |
+ double target_rms; |
|
60 |
+ double compress_factor; |
|
61 |
+ double *prev_amplification_factor; |
|
62 |
+ double *dc_correction_value; |
|
63 |
+ double *compress_threshold; |
|
64 |
+ double *fade_factors[2]; |
|
65 |
+ double *weights; |
|
66 |
+ |
|
67 |
+ int channels; |
|
68 |
+ int delay; |
|
69 |
+ |
|
70 |
+ cqueue **gain_history_original; |
|
71 |
+ cqueue **gain_history_minimum; |
|
72 |
+ cqueue **gain_history_smoothed; |
|
73 |
+} DynamicAudioNormalizerContext; |
|
74 |
+ |
|
75 |
+#define OFFSET(x) offsetof(DynamicAudioNormalizerContext, x) |
|
76 |
+#define FLAGS AV_OPT_FLAG_AUDIO_PARAM|AV_OPT_FLAG_FILTERING_PARAM |
|
77 |
+ |
|
78 |
+static const AVOption dynaudnorm_options[] = { |
|
79 |
+ { "f", "set the frame length in msec", OFFSET(frame_len_msec), AV_OPT_TYPE_INT, {.i64 = 500}, 10, 8000, FLAGS }, |
|
80 |
+ { "g", "set the filter size", OFFSET(filter_size), AV_OPT_TYPE_INT, {.i64 = 31}, 3, 301, FLAGS }, |
|
81 |
+ { "p", "set the peak value", OFFSET(peak_value), AV_OPT_TYPE_DOUBLE, {.dbl = 0.95}, 0.0, 1.0, FLAGS }, |
|
82 |
+ { "m", "set the max amplification", OFFSET(max_amplification), AV_OPT_TYPE_DOUBLE, {.dbl = 10.0}, 1.0, 100.0, FLAGS }, |
|
83 |
+ { "r", "set the target RMS", OFFSET(target_rms), AV_OPT_TYPE_DOUBLE, {.dbl = 0.0}, 0.0, 1.0, FLAGS }, |
|
84 |
+ { "n", "enable channel coupling", OFFSET(channels_coupled), AV_OPT_TYPE_INT, {.i64 = 1}, 0, 1, FLAGS }, |
|
85 |
+ { "c", "enable DC correction", OFFSET(dc_correction), AV_OPT_TYPE_INT, {.i64 = 0}, 0, 1, FLAGS }, |
|
86 |
+ { "b", "enable alternative boundary mode", OFFSET(alt_boundary_mode), AV_OPT_TYPE_INT, {.i64 = 0}, 0, 1, FLAGS }, |
|
87 |
+ { "s", "set the compress factor", OFFSET(compress_factor), AV_OPT_TYPE_DOUBLE, {.dbl = 0.0}, 0.0, 30.0, FLAGS }, |
|
88 |
+ { NULL } |
|
89 |
+}; |
|
90 |
+ |
|
91 |
+AVFILTER_DEFINE_CLASS(dynaudnorm); |
|
92 |
+ |
|
93 |
+static av_cold int init(AVFilterContext *ctx) |
|
94 |
+{ |
|
95 |
+ DynamicAudioNormalizerContext *s = ctx->priv; |
|
96 |
+ |
|
97 |
+ if (!(s->filter_size & 1)) { |
|
98 |
+ av_log(ctx, AV_LOG_ERROR, "filter size %d is invalid. Must be an odd value.\n", s->filter_size); |
|
99 |
+ return AVERROR(EINVAL); |
|
100 |
+ } |
|
101 |
+ |
|
102 |
+ return 0; |
|
103 |
+} |
|
104 |
+ |
|
105 |
+static int query_formats(AVFilterContext *ctx) |
|
106 |
+{ |
|
107 |
+ AVFilterFormats *formats; |
|
108 |
+ AVFilterChannelLayouts *layouts; |
|
109 |
+ static const enum AVSampleFormat sample_fmts[] = { |
|
110 |
+ AV_SAMPLE_FMT_DBLP, |
|
111 |
+ AV_SAMPLE_FMT_NONE |
|
112 |
+ }; |
|
113 |
+ int ret; |
|
114 |
+ |
|
115 |
+ layouts = ff_all_channel_layouts(); |
|
116 |
+ if (!layouts) |
|
117 |
+ return AVERROR(ENOMEM); |
|
118 |
+ ret = ff_set_common_channel_layouts(ctx, layouts); |
|
119 |
+ if (ret < 0) |
|
120 |
+ return ret; |
|
121 |
+ |
|
122 |
+ formats = ff_make_format_list(sample_fmts); |
|
123 |
+ if (!formats) |
|
124 |
+ return AVERROR(ENOMEM); |
|
125 |
+ ret = ff_set_common_formats(ctx, formats); |
|
126 |
+ if (ret < 0) |
|
127 |
+ return ret; |
|
128 |
+ |
|
129 |
+ formats = ff_all_samplerates(); |
|
130 |
+ if (!formats) |
|
131 |
+ return AVERROR(ENOMEM); |
|
132 |
+ return ff_set_common_samplerates(ctx, formats); |
|
133 |
+} |
|
134 |
+ |
|
135 |
+static inline int frame_size(int sample_rate, int frame_len_msec) |
|
136 |
+{ |
|
137 |
+ const int frame_size = round((double)sample_rate * (frame_len_msec / 1000.0)); |
|
138 |
+ return frame_size + (frame_size % 2); |
|
139 |
+} |
|
140 |
+ |
|
141 |
+static void precalculate_fade_factors(double *fade_factors[2], int frame_len) |
|
142 |
+{ |
|
143 |
+ const double step_size = 1.0 / frame_len; |
|
144 |
+ int pos; |
|
145 |
+ |
|
146 |
+ for (pos = 0; pos < frame_len; pos++) { |
|
147 |
+ fade_factors[0][pos] = 1.0 - (step_size * (pos + 1.0)); |
|
148 |
+ fade_factors[1][pos] = 1.0 - fade_factors[0][pos]; |
|
149 |
+ } |
|
150 |
+} |
|
151 |
+ |
|
152 |
+static cqueue *cqueue_create(int size) |
|
153 |
+{ |
|
154 |
+ cqueue *q; |
|
155 |
+ |
|
156 |
+ q = av_malloc(sizeof(cqueue)); |
|
157 |
+ if (!q) |
|
158 |
+ return NULL; |
|
159 |
+ |
|
160 |
+ q->size = size; |
|
161 |
+ q->nb_elements = 0; |
|
162 |
+ q->first = 0; |
|
163 |
+ |
|
164 |
+ q->elements = av_malloc(sizeof(double) * size); |
|
165 |
+ if (!q->elements) { |
|
166 |
+ av_free(q); |
|
167 |
+ return NULL; |
|
168 |
+ } |
|
169 |
+ |
|
170 |
+ return q; |
|
171 |
+} |
|
172 |
+ |
|
173 |
+static void cqueue_free(cqueue *q) |
|
174 |
+{ |
|
175 |
+ av_free(q->elements); |
|
176 |
+ av_free(q); |
|
177 |
+} |
|
178 |
+ |
|
179 |
+static int cqueue_size(cqueue *q) |
|
180 |
+{ |
|
181 |
+ return q->nb_elements; |
|
182 |
+} |
|
183 |
+ |
|
184 |
+static int cqueue_empty(cqueue *q) |
|
185 |
+{ |
|
186 |
+ return !q->nb_elements; |
|
187 |
+} |
|
188 |
+ |
|
189 |
+static int cqueue_enqueue(cqueue *q, double element) |
|
190 |
+{ |
|
191 |
+ int i; |
|
192 |
+ |
|
193 |
+ av_assert2(q->nb_elements |= q->size); |
|
194 |
+ |
|
195 |
+ i = (q->first + q->nb_elements) % q->size; |
|
196 |
+ q->elements[i] = element; |
|
197 |
+ q->nb_elements++; |
|
198 |
+ |
|
199 |
+ return 0; |
|
200 |
+} |
|
201 |
+ |
|
202 |
+static double cqueue_peek(cqueue *q, int index) |
|
203 |
+{ |
|
204 |
+ av_assert2(index < q->nb_elements); |
|
205 |
+ return q->elements[(q->first + index) % q->size]; |
|
206 |
+} |
|
207 |
+ |
|
208 |
+static int cqueue_dequeue(cqueue *q, double *element) |
|
209 |
+{ |
|
210 |
+ av_assert2(!cqueue_empty(q)); |
|
211 |
+ |
|
212 |
+ *element = q->elements[q->first]; |
|
213 |
+ q->first = (q->first + 1) % q->size; |
|
214 |
+ q->nb_elements--; |
|
215 |
+ |
|
216 |
+ return 0; |
|
217 |
+} |
|
218 |
+ |
|
219 |
+static int cqueue_pop(cqueue *q) |
|
220 |
+{ |
|
221 |
+ av_assert2(!cqueue_empty(q)); |
|
222 |
+ |
|
223 |
+ q->first = (q->first + 1) % q->size; |
|
224 |
+ q->nb_elements--; |
|
225 |
+ |
|
226 |
+ return 0; |
|
227 |
+} |
|
228 |
+ |
|
229 |
+static const double s_pi = 3.1415926535897932384626433832795028841971693993751058209749445923078164062862089986280348253421170679; |
|
230 |
+ |
|
231 |
+static void init_gaussian_filter(DynamicAudioNormalizerContext *s) |
|
232 |
+{ |
|
233 |
+ double total_weight = 0.0; |
|
234 |
+ const double sigma = (((s->filter_size / 2.0) - 1.0) / 3.0) + (1.0 / 3.0); |
|
235 |
+ double adjust; |
|
236 |
+ int i; |
|
237 |
+ |
|
238 |
+ // Pre-compute constants |
|
239 |
+ const int offset = s->filter_size / 2; |
|
240 |
+ const double c1 = 1.0 / (sigma * sqrt(2.0 * s_pi)); |
|
241 |
+ const double c2 = 2.0 * pow(sigma, 2.0); |
|
242 |
+ |
|
243 |
+ // Compute weights |
|
244 |
+ for (i = 0; i < s->filter_size; i++) { |
|
245 |
+ const int x = i - offset; |
|
246 |
+ |
|
247 |
+ s->weights[i] = c1 * exp(-(pow(x, 2.0) / c2)); |
|
248 |
+ total_weight += s->weights[i]; |
|
249 |
+ } |
|
250 |
+ |
|
251 |
+ // Adjust weights |
|
252 |
+ adjust = 1.0 / total_weight; |
|
253 |
+ for (i = 0; i < s->filter_size; i++) { |
|
254 |
+ s->weights[i] *= adjust; |
|
255 |
+ } |
|
256 |
+} |
|
257 |
+ |
|
258 |
+static int config_input(AVFilterLink *inlink) |
|
259 |
+{ |
|
260 |
+ AVFilterContext *ctx = inlink->dst; |
|
261 |
+ DynamicAudioNormalizerContext *s = ctx->priv; |
|
262 |
+ int c; |
|
263 |
+ |
|
264 |
+ s->frame_len = |
|
265 |
+ inlink->min_samples = |
|
266 |
+ inlink->max_samples = |
|
267 |
+ inlink->partial_buf_size = frame_size(inlink->sample_rate, s->frame_len_msec); |
|
268 |
+ av_log(ctx, AV_LOG_DEBUG, "frame len %d\n", s->frame_len); |
|
269 |
+ |
|
270 |
+ s->fade_factors[0] = av_malloc(s->frame_len * sizeof(*s->fade_factors[0])); |
|
271 |
+ s->fade_factors[1] = av_malloc(s->frame_len * sizeof(*s->fade_factors[1])); |
|
272 |
+ |
|
273 |
+ s->prev_amplification_factor = av_malloc(inlink->channels * sizeof(*s->prev_amplification_factor)); |
|
274 |
+ s->dc_correction_value = av_calloc(inlink->channels, sizeof(*s->dc_correction_value)); |
|
275 |
+ s->compress_threshold = av_calloc(inlink->channels, sizeof(*s->compress_threshold)); |
|
276 |
+ s->gain_history_original = av_calloc(inlink->channels, sizeof(*s->gain_history_original)); |
|
277 |
+ s->gain_history_minimum = av_calloc(inlink->channels, sizeof(*s->gain_history_minimum)); |
|
278 |
+ s->gain_history_smoothed = av_calloc(inlink->channels, sizeof(*s->gain_history_smoothed)); |
|
279 |
+ s->weights = av_malloc(s->filter_size * sizeof(*s->weights)); |
|
280 |
+ if (!s->prev_amplification_factor || !s->dc_correction_value || |
|
281 |
+ !s->compress_threshold || !s->fade_factors[0] || !s->fade_factors[1] || |
|
282 |
+ !s->gain_history_original || !s->gain_history_minimum || |
|
283 |
+ !s->gain_history_smoothed || !s->weights) |
|
284 |
+ return AVERROR(ENOMEM); |
|
285 |
+ |
|
286 |
+ for (c = 0; c < inlink->channels; c++) { |
|
287 |
+ s->prev_amplification_factor[c] = 1.0; |
|
288 |
+ |
|
289 |
+ s->gain_history_original[c] = cqueue_create(s->filter_size); |
|
290 |
+ s->gain_history_minimum[c] = cqueue_create(s->filter_size); |
|
291 |
+ s->gain_history_smoothed[c] = cqueue_create(s->filter_size); |
|
292 |
+ |
|
293 |
+ if (!s->gain_history_original[c] || !s->gain_history_minimum[c] || |
|
294 |
+ !s->gain_history_smoothed[c]) |
|
295 |
+ return AVERROR(ENOMEM); |
|
296 |
+ } |
|
297 |
+ |
|
298 |
+ precalculate_fade_factors(s->fade_factors, s->frame_len); |
|
299 |
+ init_gaussian_filter(s); |
|
300 |
+ |
|
301 |
+ s->channels = inlink->channels; |
|
302 |
+ s->delay = s->filter_size; |
|
303 |
+ |
|
304 |
+ return 0; |
|
305 |
+} |
|
306 |
+ |
|
307 |
+static int config_output(AVFilterLink *outlink) |
|
308 |
+{ |
|
309 |
+ outlink->flags |= FF_LINK_FLAG_REQUEST_LOOP; |
|
310 |
+ return 0; |
|
311 |
+} |
|
312 |
+ |
|
313 |
+static inline double fade(double prev, double next, int pos, |
|
314 |
+ double *fade_factors[2]) |
|
315 |
+{ |
|
316 |
+ return fade_factors[0][pos] * prev + fade_factors[1][pos] * next; |
|
317 |
+} |
|
318 |
+ |
|
319 |
+static inline double pow2(const double value) |
|
320 |
+{ |
|
321 |
+ return value * value; |
|
322 |
+} |
|
323 |
+ |
|
324 |
+static inline double bound(const double threshold, const double val) |
|
325 |
+{ |
|
326 |
+ const double CONST = 0.8862269254527580136490837416705725913987747280611935; //sqrt(PI) / 2.0 |
|
327 |
+ return erf(CONST * (val / threshold)) * threshold; |
|
328 |
+} |
|
329 |
+ |
|
330 |
+static double find_peak_magnitude(AVFrame *frame, int channel) |
|
331 |
+{ |
|
332 |
+ double max = DBL_EPSILON; |
|
333 |
+ int c, i; |
|
334 |
+ |
|
335 |
+ if (channel == -1) { |
|
336 |
+ for (c = 0; c < frame->channels; c++) { |
|
337 |
+ double *data_ptr = (double *)frame->extended_data[c]; |
|
338 |
+ |
|
339 |
+ for (i = 0; i < frame->nb_samples; i++) |
|
340 |
+ max = FFMAX(max, fabs(data_ptr[i])); |
|
341 |
+ } |
|
342 |
+ } else { |
|
343 |
+ double *data_ptr = (double *)frame->extended_data[channel]; |
|
344 |
+ |
|
345 |
+ for (i = 0; i < frame->nb_samples; i++) |
|
346 |
+ max = FFMAX(max, fabs(data_ptr[i])); |
|
347 |
+ } |
|
348 |
+ |
|
349 |
+ return max; |
|
350 |
+} |
|
351 |
+ |
|
352 |
+static double compute_frame_rms(AVFrame *frame, int channel) |
|
353 |
+{ |
|
354 |
+ double rms_value = 0.0; |
|
355 |
+ int c, i; |
|
356 |
+ |
|
357 |
+ if (channel == -1) { |
|
358 |
+ for (c = 0; c < frame->channels; c++) { |
|
359 |
+ const double *data_ptr = (double *)frame->extended_data[c]; |
|
360 |
+ |
|
361 |
+ for (i = 0; i < frame->nb_samples; i++) { |
|
362 |
+ rms_value += pow2(data_ptr[i]); |
|
363 |
+ } |
|
364 |
+ } |
|
365 |
+ |
|
366 |
+ rms_value /= frame->nb_samples * frame->channels; |
|
367 |
+ } else { |
|
368 |
+ const double *data_ptr = (double *)frame->extended_data[channel]; |
|
369 |
+ for (i = 0; i < frame->nb_samples; i++) { |
|
370 |
+ rms_value += pow2(data_ptr[i]); |
|
371 |
+ } |
|
372 |
+ |
|
373 |
+ rms_value /= frame->nb_samples; |
|
374 |
+ } |
|
375 |
+ |
|
376 |
+ return FFMAX(sqrt(rms_value), DBL_EPSILON); |
|
377 |
+} |
|
378 |
+ |
|
379 |
+static double get_max_local_gain(DynamicAudioNormalizerContext *s, AVFrame *frame, |
|
380 |
+ int channel) |
|
381 |
+{ |
|
382 |
+ const double maximum_gain = s->peak_value / find_peak_magnitude(frame, channel); |
|
383 |
+ const double rms_gain = s->target_rms > DBL_EPSILON ? (s->target_rms / compute_frame_rms(frame, channel)) : DBL_MAX; |
|
384 |
+ return bound(s->max_amplification, FFMIN(maximum_gain, rms_gain)); |
|
385 |
+} |
|
386 |
+ |
|
387 |
+static double minimum_filter(cqueue *q) |
|
388 |
+{ |
|
389 |
+ double min = DBL_MAX; |
|
390 |
+ int i; |
|
391 |
+ |
|
392 |
+ for (i = 0; i < cqueue_size(q); i++) { |
|
393 |
+ min = FFMIN(min, cqueue_peek(q, i)); |
|
394 |
+ } |
|
395 |
+ |
|
396 |
+ return min; |
|
397 |
+} |
|
398 |
+ |
|
399 |
+static double gaussian_filter(DynamicAudioNormalizerContext *s, cqueue *q) |
|
400 |
+{ |
|
401 |
+ double result = 0.0; |
|
402 |
+ int i; |
|
403 |
+ |
|
404 |
+ for (i = 0; i < cqueue_size(q); i++) { |
|
405 |
+ result += cqueue_peek(q, i) * s->weights[i]; |
|
406 |
+ } |
|
407 |
+ |
|
408 |
+ return result; |
|
409 |
+} |
|
410 |
+ |
|
411 |
+static void update_gain_history(DynamicAudioNormalizerContext *s, int channel, |
|
412 |
+ double current_gain_factor) |
|
413 |
+{ |
|
414 |
+ if (cqueue_empty(s->gain_history_original[channel]) || |
|
415 |
+ cqueue_empty(s->gain_history_minimum[channel])) { |
|
416 |
+ const int pre_fill_size = s->filter_size / 2; |
|
417 |
+ |
|
418 |
+ s->prev_amplification_factor[channel] = s->alt_boundary_mode ? current_gain_factor : 1.0; |
|
419 |
+ |
|
420 |
+ while (cqueue_size(s->gain_history_original[channel]) < pre_fill_size) { |
|
421 |
+ cqueue_enqueue(s->gain_history_original[channel], s->alt_boundary_mode ? current_gain_factor : 1.0); |
|
422 |
+ } |
|
423 |
+ |
|
424 |
+ while (cqueue_size(s->gain_history_minimum[channel]) < pre_fill_size) { |
|
425 |
+ cqueue_enqueue(s->gain_history_minimum[channel], s->alt_boundary_mode ? current_gain_factor : 1.0); |
|
426 |
+ } |
|
427 |
+ } |
|
428 |
+ |
|
429 |
+ cqueue_enqueue(s->gain_history_original[channel], current_gain_factor); |
|
430 |
+ |
|
431 |
+ while (cqueue_size(s->gain_history_original[channel]) >= s->filter_size) { |
|
432 |
+ av_assert0(cqueue_size(s->gain_history_original[channel]) == s->filter_size); |
|
433 |
+ const double minimum = minimum_filter(s->gain_history_original[channel]); |
|
434 |
+ |
|
435 |
+ cqueue_enqueue(s->gain_history_minimum[channel], minimum); |
|
436 |
+ |
|
437 |
+ cqueue_pop(s->gain_history_original[channel]); |
|
438 |
+ } |
|
439 |
+ |
|
440 |
+ while (cqueue_size(s->gain_history_minimum[channel]) >= s->filter_size) { |
|
441 |
+ av_assert0(cqueue_size(s->gain_history_minimum[channel]) == s->filter_size); |
|
442 |
+ const double smoothed = gaussian_filter(s, s->gain_history_minimum[channel]); |
|
443 |
+ |
|
444 |
+ cqueue_enqueue(s->gain_history_smoothed[channel], smoothed); |
|
445 |
+ |
|
446 |
+ cqueue_pop(s->gain_history_minimum[channel]); |
|
447 |
+ } |
|
448 |
+} |
|
449 |
+ |
|
450 |
+static inline double update_value(double new, double old, double aggressiveness) |
|
451 |
+{ |
|
452 |
+ av_assert0((aggressiveness >= 0.0) && (aggressiveness <= 1.0)); |
|
453 |
+ return aggressiveness * new + (1.0 - aggressiveness) * old; |
|
454 |
+} |
|
455 |
+ |
|
456 |
+static void perform_dc_correction(DynamicAudioNormalizerContext *s, AVFrame *frame) |
|
457 |
+{ |
|
458 |
+ const double diff = 1.0 / frame->nb_samples; |
|
459 |
+ int is_first_frame = cqueue_empty(s->gain_history_original[0]); |
|
460 |
+ int c, i; |
|
461 |
+ |
|
462 |
+ for (c = 0; c < s->channels; c++) { |
|
463 |
+ double *dst_ptr = (double *)frame->extended_data[c]; |
|
464 |
+ double current_average_value = 0.0; |
|
465 |
+ |
|
466 |
+ for (i = 0; i < frame->nb_samples; i++) |
|
467 |
+ current_average_value += dst_ptr[i] * diff; |
|
468 |
+ |
|
469 |
+ const double prev_value = is_first_frame ? current_average_value : s->dc_correction_value[c]; |
|
470 |
+ s->dc_correction_value[c] = is_first_frame ? current_average_value : update_value(current_average_value, s->dc_correction_value[c], 0.1); |
|
471 |
+ |
|
472 |
+ for (i = 0; i < frame->nb_samples; i++) { |
|
473 |
+ dst_ptr[i] -= fade(prev_value, s->dc_correction_value[c], i, s->fade_factors); |
|
474 |
+ } |
|
475 |
+ } |
|
476 |
+} |
|
477 |
+ |
|
478 |
+static double setup_compress_thresh(double threshold) |
|
479 |
+{ |
|
480 |
+ if ((threshold > DBL_EPSILON) && (threshold < (1.0 - DBL_EPSILON))) { |
|
481 |
+ double current_threshold = threshold; |
|
482 |
+ double step_size = 1.0; |
|
483 |
+ |
|
484 |
+ while (step_size > DBL_EPSILON) { |
|
485 |
+ while ((current_threshold + step_size > current_threshold) && |
|
486 |
+ (bound(current_threshold + step_size, 1.0) <= threshold)) { |
|
487 |
+ current_threshold += step_size; |
|
488 |
+ } |
|
489 |
+ |
|
490 |
+ step_size /= 2.0; |
|
491 |
+ } |
|
492 |
+ |
|
493 |
+ return current_threshold; |
|
494 |
+ } else { |
|
495 |
+ return threshold; |
|
496 |
+ } |
|
497 |
+} |
|
498 |
+ |
|
499 |
+static double compute_frame_std_dev(DynamicAudioNormalizerContext *s, |
|
500 |
+ AVFrame *frame, int channel) |
|
501 |
+{ |
|
502 |
+ double variance = 0.0; |
|
503 |
+ int i, c; |
|
504 |
+ |
|
505 |
+ if (channel == -1) { |
|
506 |
+ for (c = 0; c < s->channels; c++) { |
|
507 |
+ const double *data_ptr = (double *)frame->extended_data[c]; |
|
508 |
+ |
|
509 |
+ for (i = 0; i < frame->nb_samples; i++) { |
|
510 |
+ variance += pow2(data_ptr[i]); // Assume that MEAN is *zero* |
|
511 |
+ } |
|
512 |
+ } |
|
513 |
+ variance /= (s->channels * frame->nb_samples) - 1; |
|
514 |
+ } else { |
|
515 |
+ const double *data_ptr = (double *)frame->extended_data[channel]; |
|
516 |
+ |
|
517 |
+ for (i = 0; i < frame->nb_samples; i++) { |
|
518 |
+ variance += pow2(data_ptr[i]); // Assume that MEAN is *zero* |
|
519 |
+ } |
|
520 |
+ variance /= frame->nb_samples - 1; |
|
521 |
+ } |
|
522 |
+ |
|
523 |
+ return FFMAX(sqrt(variance), DBL_EPSILON); |
|
524 |
+} |
|
525 |
+ |
|
526 |
+static void perform_compression(DynamicAudioNormalizerContext *s, AVFrame *frame) |
|
527 |
+{ |
|
528 |
+ int is_first_frame = cqueue_empty(s->gain_history_original[0]); |
|
529 |
+ int c, i; |
|
530 |
+ |
|
531 |
+ if (s->channels_coupled) { |
|
532 |
+ const double standard_deviation = compute_frame_std_dev(s, frame, -1); |
|
533 |
+ const double current_threshold = FFMIN(1.0, s->compress_factor * standard_deviation); |
|
534 |
+ |
|
535 |
+ const double prev_value = is_first_frame ? current_threshold : s->compress_threshold[0]; |
|
536 |
+ s->compress_threshold[0] = is_first_frame ? current_threshold : update_value(current_threshold, s->compress_threshold[0], (1.0/3.0)); |
|
537 |
+ |
|
538 |
+ const double prev_actual_thresh = setup_compress_thresh(prev_value); |
|
539 |
+ const double curr_actual_thresh = setup_compress_thresh(s->compress_threshold[0]); |
|
540 |
+ |
|
541 |
+ for (c = 0; c < s->channels; c++) { |
|
542 |
+ double *const dst_ptr = (double *)frame->extended_data[c]; |
|
543 |
+ for (i = 0; i < frame->nb_samples; i++) { |
|
544 |
+ const double localThresh = fade(prev_actual_thresh, curr_actual_thresh, i, s->fade_factors); |
|
545 |
+ dst_ptr[i] = copysign(bound(localThresh, fabs(dst_ptr[i])), dst_ptr[i]); |
|
546 |
+ } |
|
547 |
+ } |
|
548 |
+ } else { |
|
549 |
+ for (c = 0; c < s->channels; c++) { |
|
550 |
+ const double standard_deviation = compute_frame_std_dev(s, frame, c); |
|
551 |
+ const double current_threshold = setup_compress_thresh(FFMIN(1.0, s->compress_factor * standard_deviation)); |
|
552 |
+ |
|
553 |
+ const double prev_value = is_first_frame ? current_threshold : s->compress_threshold[c]; |
|
554 |
+ s->compress_threshold[c] = is_first_frame ? current_threshold : update_value(current_threshold, s->compress_threshold[c], 1.0/3.0); |
|
555 |
+ |
|
556 |
+ const double prev_actual_thresh = setup_compress_thresh(prev_value); |
|
557 |
+ const double curr_actual_thresh = setup_compress_thresh(s->compress_threshold[c]); |
|
558 |
+ |
|
559 |
+ double *const dst_ptr = (double *)frame->extended_data[c]; |
|
560 |
+ for (i = 0; i < frame->nb_samples; i++) { |
|
561 |
+ const double localThresh = fade(prev_actual_thresh, curr_actual_thresh, i, s->fade_factors); |
|
562 |
+ dst_ptr[i] = copysign(bound(localThresh, fabs(dst_ptr[i])), dst_ptr[i]); |
|
563 |
+ } |
|
564 |
+ } |
|
565 |
+ } |
|
566 |
+} |
|
567 |
+ |
|
568 |
+static void analyze_frame(DynamicAudioNormalizerContext *s, AVFrame *frame) |
|
569 |
+{ |
|
570 |
+ if (s->dc_correction) { |
|
571 |
+ perform_dc_correction(s, frame); |
|
572 |
+ } |
|
573 |
+ |
|
574 |
+ if (s->compress_factor > DBL_EPSILON) { |
|
575 |
+ perform_compression(s, frame); |
|
576 |
+ } |
|
577 |
+ |
|
578 |
+ if (s->channels_coupled) { |
|
579 |
+ const double current_gain_factor = get_max_local_gain(s, frame, -1); |
|
580 |
+ int c; |
|
581 |
+ |
|
582 |
+ for (c = 0; c < s->channels; c++) |
|
583 |
+ update_gain_history(s, c, current_gain_factor); |
|
584 |
+ } else { |
|
585 |
+ int c; |
|
586 |
+ |
|
587 |
+ for (c = 0; c < s->channels; c++) |
|
588 |
+ update_gain_history(s, c, get_max_local_gain(s, frame, c)); |
|
589 |
+ } |
|
590 |
+} |
|
591 |
+ |
|
592 |
+static void amplify_frame(DynamicAudioNormalizerContext *s, AVFrame *frame) |
|
593 |
+{ |
|
594 |
+ int c, i; |
|
595 |
+ |
|
596 |
+ for (c = 0; c < s->channels; c++) { |
|
597 |
+ double *dst_ptr = (double *)frame->extended_data[c]; |
|
598 |
+ double current_amplification_factor; |
|
599 |
+ |
|
600 |
+ cqueue_dequeue(s->gain_history_smoothed[c], ¤t_amplification_factor); |
|
601 |
+ |
|
602 |
+ for (i = 0; i < frame->nb_samples; i++) { |
|
603 |
+ const double amplification_factor = fade(s->prev_amplification_factor[c], |
|
604 |
+ current_amplification_factor, i, |
|
605 |
+ s->fade_factors); |
|
606 |
+ |
|
607 |
+ dst_ptr[i] *= amplification_factor; |
|
608 |
+ |
|
609 |
+ if (fabs(dst_ptr[i]) > s->peak_value) |
|
610 |
+ dst_ptr[i] = copysign(s->peak_value, dst_ptr[i]); |
|
611 |
+ } |
|
612 |
+ |
|
613 |
+ s->prev_amplification_factor[c] = current_amplification_factor; |
|
614 |
+ } |
|
615 |
+} |
|
616 |
+ |
|
617 |
+static int filter_frame(AVFilterLink *inlink, AVFrame *in) |
|
618 |
+{ |
|
619 |
+ AVFilterContext *ctx = inlink->dst; |
|
620 |
+ DynamicAudioNormalizerContext *s = ctx->priv; |
|
621 |
+ AVFilterLink *outlink = inlink->dst->outputs[0]; |
|
622 |
+ int ret = 0; |
|
623 |
+ |
|
624 |
+ if (!cqueue_empty(s->gain_history_smoothed[0])) { |
|
625 |
+ AVFrame *out = ff_bufqueue_get(&s->queue); |
|
626 |
+ |
|
627 |
+ amplify_frame(s, out); |
|
628 |
+ ret = ff_filter_frame(outlink, out); |
|
629 |
+ } |
|
630 |
+ |
|
631 |
+ analyze_frame(s, in); |
|
632 |
+ ff_bufqueue_add(ctx, &s->queue, in); |
|
633 |
+ |
|
634 |
+ return ret; |
|
635 |
+} |
|
636 |
+ |
|
637 |
+static int flush_buffer(DynamicAudioNormalizerContext *s, AVFilterLink *inlink, |
|
638 |
+ AVFilterLink *outlink) |
|
639 |
+{ |
|
640 |
+ AVFrame *out = ff_get_audio_buffer(outlink, s->frame_len); |
|
641 |
+ int c, i; |
|
642 |
+ |
|
643 |
+ if (!out) |
|
644 |
+ return AVERROR(ENOMEM); |
|
645 |
+ |
|
646 |
+ for (c = 0; c < s->channels; c++) { |
|
647 |
+ double *dst_ptr = (double *)out->extended_data[c]; |
|
648 |
+ |
|
649 |
+ for (i = 0; i < out->nb_samples; i++) { |
|
650 |
+ dst_ptr[i] = s->alt_boundary_mode ? DBL_EPSILON : ((s->target_rms > DBL_EPSILON) ? FFMIN(s->peak_value, s->target_rms) : s->peak_value); |
|
651 |
+ if (s->dc_correction) { |
|
652 |
+ dst_ptr[i] *= ((i % 2) == 1) ? -1 : 1; |
|
653 |
+ dst_ptr[i] += s->dc_correction_value[c]; |
|
654 |
+ } |
|
655 |
+ } |
|
656 |
+ } |
|
657 |
+ |
|
658 |
+ s->delay--; |
|
659 |
+ return filter_frame(inlink, out); |
|
660 |
+} |
|
661 |
+ |
|
662 |
+static int request_frame(AVFilterLink *outlink) |
|
663 |
+{ |
|
664 |
+ AVFilterContext *ctx = outlink->src; |
|
665 |
+ DynamicAudioNormalizerContext *s = ctx->priv; |
|
666 |
+ int ret = 0; |
|
667 |
+ |
|
668 |
+ ret = ff_request_frame(ctx->inputs[0]); |
|
669 |
+ |
|
670 |
+ if (ret == AVERROR_EOF && !ctx->is_disabled && s->delay) |
|
671 |
+ ret = flush_buffer(s, ctx->inputs[0], outlink); |
|
672 |
+ |
|
673 |
+ return ret; |
|
674 |
+} |
|
675 |
+ |
|
676 |
+static av_cold void uninit(AVFilterContext *ctx) |
|
677 |
+{ |
|
678 |
+ DynamicAudioNormalizerContext *s = ctx->priv; |
|
679 |
+ int c; |
|
680 |
+ |
|
681 |
+ av_freep(&s->prev_amplification_factor); |
|
682 |
+ av_freep(&s->dc_correction_value); |
|
683 |
+ av_freep(&s->compress_threshold); |
|
684 |
+ av_freep(&s->fade_factors[0]); |
|
685 |
+ av_freep(&s->fade_factors[1]); |
|
686 |
+ |
|
687 |
+ for (c = 0; c < s->channels; c++) { |
|
688 |
+ cqueue_free(s->gain_history_original[c]); |
|
689 |
+ cqueue_free(s->gain_history_minimum[c]); |
|
690 |
+ cqueue_free(s->gain_history_smoothed[c]); |
|
691 |
+ } |
|
692 |
+ |
|
693 |
+ av_freep(&s->gain_history_original); |
|
694 |
+ av_freep(&s->gain_history_minimum); |
|
695 |
+ av_freep(&s->gain_history_smoothed); |
|
696 |
+ |
|
697 |
+ av_freep(&s->weights); |
|
698 |
+ |
|
699 |
+ ff_bufqueue_discard_all(&s->queue); |
|
700 |
+} |
|
701 |
+ |
|
702 |
+static const AVFilterPad avfilter_af_dynaudnorm_inputs[] = { |
|
703 |
+ { |
|
704 |
+ .name = "default", |
|
705 |
+ .type = AVMEDIA_TYPE_AUDIO, |
|
706 |
+ .filter_frame = filter_frame, |
|
707 |
+ .config_props = config_input, |
|
708 |
+ .needs_writable = 1, |
|
709 |
+ }, |
|
710 |
+ { NULL } |
|
711 |
+}; |
|
712 |
+ |
|
713 |
+static const AVFilterPad avfilter_af_dynaudnorm_outputs[] = { |
|
714 |
+ { |
|
715 |
+ .name = "default", |
|
716 |
+ .type = AVMEDIA_TYPE_AUDIO, |
|
717 |
+ .config_props = config_output, |
|
718 |
+ .request_frame = request_frame, |
|
719 |
+ }, |
|
720 |
+ { NULL } |
|
721 |
+}; |
|
722 |
+ |
|
723 |
+AVFilter ff_af_dynaudnorm = { |
|
724 |
+ .name = "dynaudnorm", |
|
725 |
+ .description = NULL_IF_CONFIG_SMALL("Dynamic Audio Normalizer."), |
|
726 |
+ .query_formats = query_formats, |
|
727 |
+ .priv_size = sizeof(DynamicAudioNormalizerContext), |
|
728 |
+ .init = init, |
|
729 |
+ .uninit = uninit, |
|
730 |
+ .inputs = avfilter_af_dynaudnorm_inputs, |
|
731 |
+ .outputs = avfilter_af_dynaudnorm_outputs, |
|
732 |
+ .priv_class = &dynaudnorm_class, |
|
733 |
+}; |
... | ... |
@@ -83,6 +83,7 @@ void avfilter_register_all(void) |
83 | 83 |
REGISTER_FILTER(CHORUS, chorus, af); |
84 | 84 |
REGISTER_FILTER(COMPAND, compand, af); |
85 | 85 |
REGISTER_FILTER(DCSHIFT, dcshift, af); |
86 |
+ REGISTER_FILTER(DYNAUDNORM, dynaudnorm, af); |
|
86 | 87 |
REGISTER_FILTER(EARWAX, earwax, af); |
87 | 88 |
REGISTER_FILTER(EBUR128, ebur128, af); |
88 | 89 |
REGISTER_FILTER(EQUALIZER, equalizer, af); |