FFmpeg
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
vf_mpdecimate.c
Go to the documentation of this file.
1 /*
2  * Copyright (c) 2003 Rich Felker
3  * Copyright (c) 2012 Stefano Sabatini
4  *
5  * This file is part of FFmpeg.
6  *
7  * FFmpeg is free software; you can redistribute it and/or modify
8  * it under the terms of the GNU General Public License as published by
9  * the Free Software Foundation; either version 2 of the License, or
10  * (at your option) any later version.
11  *
12  * FFmpeg is distributed in the hope that it will be useful,
13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15  * GNU General Public License for more details.
16  *
17  * You should have received a copy of the GNU General Public License along
18  * with FFmpeg; if not, write to the Free Software Foundation, Inc.,
19  * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
20  */
21 
22 /**
23  * @file mpdecimate filter, ported from libmpcodecs/vf_decimate.c by
24  * Rich Felker.
25  */
26 
27 #include "libavutil/opt.h"
28 #include "libavutil/pixdesc.h"
29 #include "libavutil/timestamp.h"
30 #include "libavcodec/dsputil.h"
31 #include "avfilter.h"
32 #include "internal.h"
33 #include "formats.h"
34 #include "video.h"
35 
36 typedef struct {
37  const AVClass *class;
38  int lo, hi; ///< lower and higher threshold number of differences
39  ///< values for 8x8 blocks
40 
41  float frac; ///< threshold of changed pixels over the total fraction
42 
43  int max_drop_count; ///< if positive: maximum number of sequential frames to drop
44  ///< if negative: minimum number of frames between two drops
45 
46  int drop_count; ///< if positive: number of frames sequentially dropped
47  ///< if negative: number of sequential frames which were not dropped
48 
49  int hsub, vsub; ///< chroma subsampling values
50  AVFrame *ref; ///< reference picture
51  DSPContext dspctx; ///< context providing optimized diff routines
52  AVCodecContext *avctx; ///< codec context required for the DSPContext
54 
55 #define OFFSET(x) offsetof(DecimateContext, x)
56 #define FLAGS AV_OPT_FLAG_VIDEO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
57 
58 static const AVOption mpdecimate_options[] = {
59  { "max", "set the maximum number of consecutive dropped frames (positive), or the minimum interval between dropped frames (negative)",
60  OFFSET(max_drop_count), AV_OPT_TYPE_INT, {.i64=0}, INT_MIN, INT_MAX, FLAGS },
61  { "hi", "set high dropping threshold", OFFSET(hi), AV_OPT_TYPE_INT, {.i64=64*12}, INT_MIN, INT_MAX, FLAGS },
62  { "lo", "set low dropping threshold", OFFSET(lo), AV_OPT_TYPE_INT, {.i64=64*5}, INT_MIN, INT_MAX, FLAGS },
63  { "frac", "set fraction dropping threshold", OFFSET(frac), AV_OPT_TYPE_FLOAT, {.dbl=0.33}, 0, 1, FLAGS },
64  { NULL }
65 };
66 
67 AVFILTER_DEFINE_CLASS(mpdecimate);
68 
69 /**
70  * Return 1 if the two planes are different, 0 otherwise.
71  */
72 static int diff_planes(AVFilterContext *ctx,
73  uint8_t *cur, uint8_t *ref, int linesize,
74  int w, int h)
75 {
76  DecimateContext *decimate = ctx->priv;
77  DSPContext *dspctx = &decimate->dspctx;
78 
79  int x, y;
80  int d, c = 0;
81  int t = (w/16)*(h/16)*decimate->frac;
82  int16_t block[8*8];
83 
84  /* compute difference for blocks of 8x8 bytes */
85  for (y = 0; y < h-7; y += 4) {
86  for (x = 8; x < w-7; x += 4) {
87  dspctx->diff_pixels(block,
88  cur+x+y*linesize,
89  ref+x+y*linesize, linesize);
90  d = dspctx->sum_abs_dctelem(block);
91  if (d > decimate->hi)
92  return 1;
93  if (d > decimate->lo) {
94  c++;
95  if (c > t)
96  return 1;
97  }
98  }
99  }
100  return 0;
101 }
102 
103 /**
104  * Tell if the frame should be decimated, for example if it is no much
105  * different with respect to the reference frame ref.
106  */
108  AVFrame *cur, AVFrame *ref)
109 {
110  DecimateContext *decimate = ctx->priv;
111  int plane;
112 
113  if (decimate->max_drop_count > 0 &&
114  decimate->drop_count >= decimate->max_drop_count)
115  return 0;
116  if (decimate->max_drop_count < 0 &&
117  (decimate->drop_count-1) > decimate->max_drop_count)
118  return 0;
119 
120  for (plane = 0; ref->data[plane] && ref->linesize[plane]; plane++) {
121  int vsub = plane == 1 || plane == 2 ? decimate->vsub : 0;
122  int hsub = plane == 1 || plane == 2 ? decimate->hsub : 0;
123  if (diff_planes(ctx,
124  cur->data[plane], ref->data[plane], ref->linesize[plane],
125  FF_CEIL_RSHIFT(ref->width, hsub),
126  FF_CEIL_RSHIFT(ref->height, vsub)))
127  return 0;
128  }
129 
130  return 1;
131 }
132 
133 static av_cold int init(AVFilterContext *ctx)
134 {
135  DecimateContext *decimate = ctx->priv;
136 
137  av_log(ctx, AV_LOG_VERBOSE, "max_drop_count:%d hi:%d lo:%d frac:%f\n",
138  decimate->max_drop_count, decimate->hi, decimate->lo, decimate->frac);
139 
140  decimate->avctx = avcodec_alloc_context3(NULL);
141  if (!decimate->avctx)
142  return AVERROR(ENOMEM);
143  avpriv_dsputil_init(&decimate->dspctx, decimate->avctx);
144 
145  return 0;
146 }
147 
148 static av_cold void uninit(AVFilterContext *ctx)
149 {
150  DecimateContext *decimate = ctx->priv;
151  av_frame_free(&decimate->ref);
152  if (decimate->avctx) {
153  avcodec_close(decimate->avctx);
154  av_freep(&decimate->avctx);
155  }
156 }
157 
159 {
160  static const enum AVPixelFormat pix_fmts[] = {
168  };
169 
171 
172  return 0;
173 }
174 
175 static int config_input(AVFilterLink *inlink)
176 {
177  AVFilterContext *ctx = inlink->dst;
178  DecimateContext *decimate = ctx->priv;
179  const AVPixFmtDescriptor *pix_desc = av_pix_fmt_desc_get(inlink->format);
180  decimate->hsub = pix_desc->log2_chroma_w;
181  decimate->vsub = pix_desc->log2_chroma_h;
182 
183  return 0;
184 }
185 
186 static int filter_frame(AVFilterLink *inlink, AVFrame *cur)
187 {
188  DecimateContext *decimate = inlink->dst->priv;
189  AVFilterLink *outlink = inlink->dst->outputs[0];
190  int ret;
191 
192  if (decimate->ref && decimate_frame(inlink->dst, cur, decimate->ref)) {
193  decimate->drop_count = FFMAX(1, decimate->drop_count+1);
194  } else {
195  av_frame_free(&decimate->ref);
196  decimate->ref = cur;
197  decimate->drop_count = FFMIN(-1, decimate->drop_count-1);
198 
199  if (ret = ff_filter_frame(outlink, av_frame_clone(cur)) < 0)
200  return ret;
201  }
202 
203  av_log(inlink->dst, AV_LOG_DEBUG,
204  "%s pts:%s pts_time:%s drop_count:%d\n",
205  decimate->drop_count > 0 ? "drop" : "keep",
206  av_ts2str(cur->pts), av_ts2timestr(cur->pts, &inlink->time_base),
207  decimate->drop_count);
208 
209  if (decimate->drop_count > 0)
210  av_frame_free(&cur);
211 
212  return 0;
213 }
214 
215 static int request_frame(AVFilterLink *outlink)
216 {
217  DecimateContext *decimate = outlink->src->priv;
218  AVFilterLink *inlink = outlink->src->inputs[0];
219  int ret;
220 
221  do {
222  ret = ff_request_frame(inlink);
223  } while (decimate->drop_count > 0 && ret >= 0);
224 
225  return ret;
226 }
227 
228 static const AVFilterPad mpdecimate_inputs[] = {
229  {
230  .name = "default",
231  .type = AVMEDIA_TYPE_VIDEO,
232  .get_video_buffer = ff_null_get_video_buffer,
233  .config_props = config_input,
234  .filter_frame = filter_frame,
235  },
236  { NULL }
237 };
238 
239 static const AVFilterPad mpdecimate_outputs[] = {
240  {
241  .name = "default",
242  .type = AVMEDIA_TYPE_VIDEO,
243  .request_frame = request_frame,
244  },
245  { NULL }
246 };
247 
249  .name = "mpdecimate",
250  .description = NULL_IF_CONFIG_SMALL("Remove near-duplicate frames."),
251  .init = init,
252  .uninit = uninit,
253 
254  .priv_size = sizeof(DecimateContext),
256  .inputs = mpdecimate_inputs,
257  .outputs = mpdecimate_outputs,
258  .priv_class = &mpdecimate_class,
259 };