1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
|
/*
* Copyright (c) 2003 Rich Felker
* Copyright (c) 2012 Stefano Sabatini
*
* This file is part of FFmpeg.
*
* FFmpeg is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; either version 2 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 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 mpdecimate filter, ported from libmpcodecs/vf_decimate.c by
* Rich Felker.
*/
#include "libavutil/opt.h"
#include "libavutil/pixdesc.h"
#include "libavutil/pixelutils.h"
#include "libavutil/timestamp.h"
#include "avfilter.h"
#include "filters.h"
#include "video.h"
typedef struct DecimateContext {
const AVClass *class;
int lo, hi; ///< lower and higher threshold number of differences
///< values for 8x8 blocks
float frac; ///< threshold of changed pixels over the total fraction
int max_drop_count; ///< if positive: maximum number of sequential frames to drop
///< if negative: minimum number of frames between two drops
int drop_count; ///< if positive: number of frames sequentially dropped
///< if negative: number of sequential frames which were not dropped
int max_keep_count; ///< number of similar frames to ignore before to start dropping them
int keep_count; ///< number of similar frames already ignored
int hsub, vsub; ///< chroma subsampling values
AVFrame *ref; ///< reference picture
av_pixelutils_sad_fn sad; ///< sum of absolute difference function
} DecimateContext;
#define OFFSET(x) offsetof(DecimateContext, x)
#define FLAGS AV_OPT_FLAG_VIDEO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
static const AVOption mpdecimate_options[] = {
{ "max", "set the maximum number of consecutive dropped frames (positive), or the minimum interval between dropped frames (negative)",
OFFSET(max_drop_count), AV_OPT_TYPE_INT, {.i64=0}, INT_MIN, INT_MAX, FLAGS },
{ "keep", "set the number of similar consecutive frames to be kept before starting to drop similar frames",
OFFSET(max_keep_count), AV_OPT_TYPE_INT, {.i64=0}, 0, INT_MAX, FLAGS },
{ "hi", "set high dropping threshold", OFFSET(hi), AV_OPT_TYPE_INT, {.i64=64*12}, INT_MIN, INT_MAX, FLAGS },
{ "lo", "set low dropping threshold", OFFSET(lo), AV_OPT_TYPE_INT, {.i64=64*5}, INT_MIN, INT_MAX, FLAGS },
{ "frac", "set fraction dropping threshold", OFFSET(frac), AV_OPT_TYPE_FLOAT, {.dbl=0.33}, 0, 1, FLAGS },
{ NULL }
};
AVFILTER_DEFINE_CLASS(mpdecimate);
/**
* Return 1 if the two planes are different, 0 otherwise.
*/
static int diff_planes(AVFilterContext *ctx,
uint8_t *cur, int cur_linesize,
uint8_t *ref, int ref_linesize,
int w, int h)
{
DecimateContext *decimate = ctx->priv;
int x, y;
int d, c = 0;
int t = (w/16)*(h/16)*decimate->frac;
/* compute difference for blocks of 8x8 bytes */
for (y = 0; y < h-7; y += 4) {
for (x = 8; x < w-7; x += 4) {
d = decimate->sad(cur + y*cur_linesize + x, cur_linesize,
ref + y*ref_linesize + x, ref_linesize);
if (d > decimate->hi) {
av_log(ctx, AV_LOG_DEBUG, "%d>=hi ", d);
return 1;
}
if (d > decimate->lo) {
c++;
if (c > t) {
av_log(ctx, AV_LOG_DEBUG, "lo:%d>=%d ", c, t);
return 1;
}
}
}
}
av_log(ctx, AV_LOG_DEBUG, "lo:%d<%d ", c, t);
return 0;
}
/**
* Tell if the frame should be decimated, for example if it is no much
* different with respect to the reference frame ref.
*/
static int decimate_frame(AVFilterContext *ctx,
AVFrame *cur, AVFrame *ref)
{
DecimateContext *decimate = ctx->priv;
int plane;
if (decimate->max_keep_count > 0 &&
decimate->keep_count > -1 &&
decimate->keep_count < decimate->max_keep_count) {
decimate->keep_count++;
return 0;
}
if (decimate->max_drop_count > 0 &&
decimate->drop_count >= decimate->max_drop_count)
return 0;
if (decimate->max_drop_count < 0 &&
(decimate->drop_count-1) > decimate->max_drop_count)
return 0;
for (plane = 0; ref->data[plane] && ref->linesize[plane]; plane++) {
/* use 8x8 SAD even on subsampled planes. The blocks won't match up with
* luma blocks, but hopefully nobody is depending on this to catch
* localized chroma changes that wouldn't exceed the thresholds when
* diluted by using what's effectively a larger block size.
*/
int vsub = plane == 1 || plane == 2 ? decimate->vsub : 0;
int hsub = plane == 1 || plane == 2 ? decimate->hsub : 0;
if (diff_planes(ctx,
cur->data[plane], cur->linesize[plane],
ref->data[plane], ref->linesize[plane],
AV_CEIL_RSHIFT(ref->width, hsub),
AV_CEIL_RSHIFT(ref->height, vsub)))
return 0;
}
return 1;
}
static av_cold int init(AVFilterContext *ctx)
{
DecimateContext *decimate = ctx->priv;
decimate->sad = av_pixelutils_get_sad_fn(3, 3, 0, ctx); // 8x8, not aligned on blocksize
if (!decimate->sad)
return AVERROR(EINVAL);
av_log(ctx, AV_LOG_VERBOSE, "max_drop_count:%d hi:%d lo:%d frac:%f\n",
decimate->max_drop_count, decimate->hi, decimate->lo, decimate->frac);
return 0;
}
static av_cold void uninit(AVFilterContext *ctx)
{
DecimateContext *decimate = ctx->priv;
av_frame_free(&decimate->ref);
}
static const enum AVPixelFormat pix_fmts[] = {
AV_PIX_FMT_YUV444P, AV_PIX_FMT_YUV422P,
AV_PIX_FMT_YUV420P, AV_PIX_FMT_YUV411P,
AV_PIX_FMT_YUV410P, AV_PIX_FMT_YUV440P,
AV_PIX_FMT_YUVJ444P, AV_PIX_FMT_YUVJ422P,
AV_PIX_FMT_YUVJ420P, AV_PIX_FMT_YUVJ440P,
AV_PIX_FMT_YUVA420P,
AV_PIX_FMT_GBRP,
AV_PIX_FMT_YUVA444P,
AV_PIX_FMT_YUVA422P,
AV_PIX_FMT_NONE
};
static int config_input(AVFilterLink *inlink)
{
AVFilterContext *ctx = inlink->dst;
DecimateContext *decimate = ctx->priv;
const AVPixFmtDescriptor *pix_desc = av_pix_fmt_desc_get(inlink->format);
decimate->hsub = pix_desc->log2_chroma_w;
decimate->vsub = pix_desc->log2_chroma_h;
return 0;
}
static int filter_frame(AVFilterLink *inlink, AVFrame *cur)
{
DecimateContext *decimate = inlink->dst->priv;
AVFilterLink *outlink = inlink->dst->outputs[0];
int ret;
if (decimate->ref && decimate_frame(inlink->dst, cur, decimate->ref)) {
decimate->drop_count = FFMAX(1, decimate->drop_count+1);
decimate->keep_count = -1; // do not keep any more frames until non-similar frames are detected
} else {
av_frame_free(&decimate->ref);
decimate->ref = cur;
decimate->drop_count = FFMIN(-1, decimate->drop_count-1);
if (decimate->keep_count < 0) // re-enable counting similiar frames to ignore before dropping
decimate->keep_count = 0;
if ((ret = ff_filter_frame(outlink, av_frame_clone(cur))) < 0)
return ret;
}
av_log(inlink->dst, AV_LOG_DEBUG,
"%s pts:%s pts_time:%s drop_count:%d keep_count:%d\n",
decimate->drop_count > 0 ? "drop" : "keep",
av_ts2str(cur->pts), av_ts2timestr(cur->pts, &inlink->time_base),
decimate->drop_count,
decimate->keep_count);
if (decimate->drop_count > 0)
av_frame_free(&cur);
return 0;
}
static const AVFilterPad mpdecimate_inputs[] = {
{
.name = "default",
.type = AVMEDIA_TYPE_VIDEO,
.config_props = config_input,
.filter_frame = filter_frame,
},
};
const AVFilter ff_vf_mpdecimate = {
.name = "mpdecimate",
.description = NULL_IF_CONFIG_SMALL("Remove near-duplicate frames."),
.init = init,
.uninit = uninit,
.priv_size = sizeof(DecimateContext),
.priv_class = &mpdecimate_class,
FILTER_INPUTS(mpdecimate_inputs),
FILTER_OUTPUTS(ff_video_default_filterpad),
FILTER_PIXFMTS_ARRAY(pix_fmts),
};
|
