FFmpeg
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/pixelutils.h"
30 #include "libavutil/timestamp.h"
31 #include "avfilter.h"
32 #include "internal.h"
33 #include "video.h"
34 
35 typedef struct DecimateContext {
36  const AVClass *class;
37  int lo, hi; ///< lower and higher threshold number of differences
38  ///< values for 8x8 blocks
39 
40  float frac; ///< threshold of changed pixels over the total fraction
41 
42  int max_drop_count; ///< if positive: maximum number of sequential frames to drop
43  ///< if negative: minimum number of frames between two drops
44 
45  int drop_count; ///< if positive: number of frames sequentially dropped
46  ///< if negative: number of sequential frames which were not dropped
47 
48  int max_keep_count; ///< number of similar frames to ignore before to start dropping them
49  int keep_count; ///< number of similar frames already ignored
50 
51  int hsub, vsub; ///< chroma subsampling values
52  AVFrame *ref; ///< reference picture
53  av_pixelutils_sad_fn sad; ///< sum of absolute difference function
55 
56 #define OFFSET(x) offsetof(DecimateContext, x)
57 #define FLAGS AV_OPT_FLAG_VIDEO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
58 
59 static const AVOption mpdecimate_options[] = {
60  { "max", "set the maximum number of consecutive dropped frames (positive), or the minimum interval between dropped frames (negative)",
61  OFFSET(max_drop_count), AV_OPT_TYPE_INT, {.i64=0}, INT_MIN, INT_MAX, FLAGS },
62  { "keep", "set the number of similar consecutive frames to be kept before starting to drop similar frames",
63  OFFSET(max_keep_count), AV_OPT_TYPE_INT, {.i64=0}, 0, INT_MAX, FLAGS },
64  { "hi", "set high dropping threshold", OFFSET(hi), AV_OPT_TYPE_INT, {.i64=64*12}, INT_MIN, INT_MAX, FLAGS },
65  { "lo", "set low dropping threshold", OFFSET(lo), AV_OPT_TYPE_INT, {.i64=64*5}, INT_MIN, INT_MAX, FLAGS },
66  { "frac", "set fraction dropping threshold", OFFSET(frac), AV_OPT_TYPE_FLOAT, {.dbl=0.33}, 0, 1, FLAGS },
67  { NULL }
68 };
69 
70 AVFILTER_DEFINE_CLASS(mpdecimate);
71 
72 /**
73  * Return 1 if the two planes are different, 0 otherwise.
74  */
76  uint8_t *cur, int cur_linesize,
77  uint8_t *ref, int ref_linesize,
78  int w, int h)
79 {
80  DecimateContext *decimate = ctx->priv;
81 
82  int x, y;
83  int d, c = 0;
84  int t = (w/16)*(h/16)*decimate->frac;
85 
86  /* compute difference for blocks of 8x8 bytes */
87  for (y = 0; y < h-7; y += 4) {
88  for (x = 8; x < w-7; x += 4) {
89  d = decimate->sad(cur + y*cur_linesize + x, cur_linesize,
90  ref + y*ref_linesize + x, ref_linesize);
91  if (d > decimate->hi) {
92  av_log(ctx, AV_LOG_DEBUG, "%d>=hi ", d);
93  return 1;
94  }
95  if (d > decimate->lo) {
96  c++;
97  if (c > t) {
98  av_log(ctx, AV_LOG_DEBUG, "lo:%d>=%d ", c, t);
99  return 1;
100  }
101  }
102  }
103  }
104 
105  av_log(ctx, AV_LOG_DEBUG, "lo:%d<%d ", c, t);
106  return 0;
107 }
108 
109 /**
110  * Tell if the frame should be decimated, for example if it is no much
111  * different with respect to the reference frame ref.
112  */
114  AVFrame *cur, AVFrame *ref)
115 {
116  DecimateContext *decimate = ctx->priv;
117  int plane;
118 
119  if (decimate->max_keep_count > 0 &&
120  decimate->keep_count > -1 &&
121  decimate->keep_count < decimate->max_keep_count) {
122  decimate->keep_count++;
123  return 0;
124  }
125  if (decimate->max_drop_count > 0 &&
126  decimate->drop_count >= decimate->max_drop_count)
127  return 0;
128  if (decimate->max_drop_count < 0 &&
129  (decimate->drop_count-1) > decimate->max_drop_count)
130  return 0;
131 
132  for (plane = 0; ref->data[plane] && ref->linesize[plane]; plane++) {
133  /* use 8x8 SAD even on subsampled planes. The blocks won't match up with
134  * luma blocks, but hopefully nobody is depending on this to catch
135  * localized chroma changes that wouldn't exceed the thresholds when
136  * diluted by using what's effectively a larger block size.
137  */
138  int vsub = plane == 1 || plane == 2 ? decimate->vsub : 0;
139  int hsub = plane == 1 || plane == 2 ? decimate->hsub : 0;
140  if (diff_planes(ctx,
141  cur->data[plane], cur->linesize[plane],
142  ref->data[plane], ref->linesize[plane],
143  AV_CEIL_RSHIFT(ref->width, hsub),
144  AV_CEIL_RSHIFT(ref->height, vsub)))
145  return 0;
146  }
147 
148  return 1;
149 }
150 
152 {
153  DecimateContext *decimate = ctx->priv;
154 
155  decimate->sad = av_pixelutils_get_sad_fn(3, 3, 0, ctx); // 8x8, not aligned on blocksize
156  if (!decimate->sad)
157  return AVERROR(EINVAL);
158 
159  av_log(ctx, AV_LOG_VERBOSE, "max_drop_count:%d hi:%d lo:%d frac:%f\n",
160  decimate->max_drop_count, decimate->hi, decimate->lo, decimate->frac);
161 
162  return 0;
163 }
164 
166 {
167  DecimateContext *decimate = ctx->priv;
168  av_frame_free(&decimate->ref);
169 }
170 
171 static const enum AVPixelFormat pix_fmts[] = {
178 
180 
183 
185 };
186 
188 {
189  AVFilterContext *ctx = inlink->dst;
190  DecimateContext *decimate = ctx->priv;
191  const AVPixFmtDescriptor *pix_desc = av_pix_fmt_desc_get(inlink->format);
192  decimate->hsub = pix_desc->log2_chroma_w;
193  decimate->vsub = pix_desc->log2_chroma_h;
194 
195  return 0;
196 }
197 
199 {
200  DecimateContext *decimate = inlink->dst->priv;
201  AVFilterLink *outlink = inlink->dst->outputs[0];
202  int ret;
203 
204  if (decimate->ref && decimate_frame(inlink->dst, cur, decimate->ref)) {
205  decimate->drop_count = FFMAX(1, decimate->drop_count+1);
206  decimate->keep_count = -1; // do not keep any more frames until non-similar frames are detected
207  } else {
208  av_frame_free(&decimate->ref);
209  decimate->ref = cur;
210  decimate->drop_count = FFMIN(-1, decimate->drop_count-1);
211  if (decimate->keep_count < 0) // re-enable counting similiar frames to ignore before dropping
212  decimate->keep_count = 0;
213 
214  if ((ret = ff_filter_frame(outlink, av_frame_clone(cur))) < 0)
215  return ret;
216  }
217 
218  av_log(inlink->dst, AV_LOG_DEBUG,
219  "%s pts:%s pts_time:%s drop_count:%d keep_count:%d\n",
220  decimate->drop_count > 0 ? "drop" : "keep",
221  av_ts2str(cur->pts), av_ts2timestr(cur->pts, &inlink->time_base),
222  decimate->drop_count,
223  decimate->keep_count);
224 
225  if (decimate->drop_count > 0)
226  av_frame_free(&cur);
227 
228  return 0;
229 }
230 
231 static const AVFilterPad mpdecimate_inputs[] = {
232  {
233  .name = "default",
234  .type = AVMEDIA_TYPE_VIDEO,
235  .config_props = config_input,
236  .filter_frame = filter_frame,
237  },
238 };
239 
241  .name = "mpdecimate",
242  .description = NULL_IF_CONFIG_SMALL("Remove near-duplicate frames."),
243  .init = init,
244  .uninit = uninit,
245  .priv_size = sizeof(DecimateContext),
246  .priv_class = &mpdecimate_class,
250 };
DecimateContext::lo
int lo
Definition: vf_mpdecimate.c:37
AVPixelFormat
AVPixelFormat
Pixel format.
Definition: pixfmt.h:71
AVERROR
Filter the word “frame” indicates either a video frame or a group of audio as stored in an AVFrame structure Format for each input and each output the list of supported formats For video that means pixel format For audio that means channel sample they are references to shared objects When the negotiation mechanism computes the intersection of the formats supported at each end of a all references to both lists are replaced with a reference to the intersection And when a single format is eventually chosen for a link amongst the remaining all references to the list are updated That means that if a filter requires that its input and output have the same format amongst a supported all it has to do is use a reference to the same list of formats query_formats can leave some formats unset and return AVERROR(EAGAIN) to cause the negotiation mechanism toagain later. That can be used by filters with complex requirements to use the format negotiated on one link to set the formats supported on another. Frame references ownership and permissions
opt.h
ff_filter_frame
int ff_filter_frame(AVFilterLink *link, AVFrame *frame)
Send a frame of data to the next filter.
Definition: avfilter.c:1018
av_pix_fmt_desc_get
const AVPixFmtDescriptor * av_pix_fmt_desc_get(enum AVPixelFormat pix_fmt)
Definition: pixdesc.c:2962
FILTER_PIXFMTS_ARRAY
#define FILTER_PIXFMTS_ARRAY(array)
Definition: internal.h:162
inlink
The exact code depends on how similar the blocks are and how related they are to the and needs to apply these operations to the correct inlink or outlink if there are several Macros are available to factor that when no extra processing is inlink
Definition: filter_design.txt:212
av_frame_free
void av_frame_free(AVFrame **frame)
Free the frame and any dynamically allocated objects in it, e.g.
Definition: frame.c:88
av_pixelutils_sad_fn
int(* av_pixelutils_sad_fn)(const uint8_t *src1, ptrdiff_t stride1, const uint8_t *src2, ptrdiff_t stride2)
Sum of abs(src1[x] - src2[x])
Definition: pixelutils.h:28
AVFrame
This structure describes decoded (raw) audio or video data.
Definition: frame.h:340
pixdesc.h
AVFrame::pts
int64_t pts
Presentation timestamp in time_base units (time when frame should be shown to user).
Definition: frame.h:452
w
uint8_t w
Definition: llviddspenc.c:38
AVOption
AVOption.
Definition: opt.h:346
AV_LOG_VERBOSE
#define AV_LOG_VERBOSE
Detailed information.
Definition: log.h:196
DecimateContext::vsub
int vsub
chroma subsampling values
Definition: vf_decimate.c:50
AV_PIX_FMT_YUV440P
@ AV_PIX_FMT_YUV440P
planar YUV 4:4:0 (1 Cr & Cb sample per 1x2 Y samples)
Definition: pixfmt.h:106
FFMAX
#define FFMAX(a, b)
Definition: macros.h:47
AVFilter::name
const char * name
Filter name.
Definition: avfilter.h:170
video.h
DecimateContext::hsub
int hsub
Definition: vf_decimate.c:50
AVFrame::data
uint8_t * data[AV_NUM_DATA_POINTERS]
pointer to the picture/channel planes.
Definition: frame.h:361
hsub
static void hsub(htype *dst, const htype *src, int bins)
Definition: vf_median.c:73
DecimateContext::drop_count
int drop_count
if positive: number of frames sequentially dropped if negative: number of sequential frames which wer...
Definition: vf_mpdecimate.c:45
config_input
static int config_input(AVFilterLink *inlink)
Definition: vf_mpdecimate.c:187
DecimateContext::keep_count
int keep_count
number of similar frames already ignored
Definition: vf_mpdecimate.c:49
AVFilterPad
A filter pad used for either input or output.
Definition: internal.h:33
av_cold
#define av_cold
Definition: attributes.h:90
ff_video_default_filterpad
const AVFilterPad ff_video_default_filterpad[1]
An AVFilterPad array whose only entry has name "default" and is of type AVMEDIA_TYPE_VIDEO.
Definition: video.c:37
filter_frame
static int filter_frame(AVFilterLink *inlink, AVFrame *cur)
Definition: vf_mpdecimate.c:198
AV_PIX_FMT_YUVJ422P
@ AV_PIX_FMT_YUVJ422P
planar YUV 4:2:2, 16bpp, full scale (JPEG), deprecated in favor of AV_PIX_FMT_YUV422P and setting col...
Definition: pixfmt.h:86
AV_PIX_FMT_YUVA420P
@ AV_PIX_FMT_YUVA420P
planar YUV 4:2:0, 20bpp, (1 Cr & Cb sample per 2x2 Y & A samples)
Definition: pixfmt.h:108
AV_CEIL_RSHIFT
#define AV_CEIL_RSHIFT(a, b)
Definition: common.h:58
AV_LOG_DEBUG
#define AV_LOG_DEBUG
Stuff which is only useful for libav* developers.
Definition: log.h:201
FLAGS
#define FLAGS
Definition: vf_mpdecimate.c:57
ctx
AVFormatContext * ctx
Definition: movenc.c:48
pix_fmts
static enum AVPixelFormat pix_fmts[]
Definition: vf_mpdecimate.c:171
av_frame_clone
AVFrame * av_frame_clone(const AVFrame *src)
Create a new frame that references the same data as src.
Definition: frame.c:521
AVPixFmtDescriptor::log2_chroma_w
uint8_t log2_chroma_w
Amount to shift the luma width right to find the chroma width.
Definition: pixdesc.h:80
AV_PIX_FMT_YUV420P
@ AV_PIX_FMT_YUV420P
planar YUV 4:2:0, 12bpp, (1 Cr & Cb sample per 2x2 Y samples)
Definition: pixfmt.h:73
FILTER_INPUTS
#define FILTER_INPUTS(array)
Definition: internal.h:182
AV_PIX_FMT_YUVJ444P
@ AV_PIX_FMT_YUVJ444P
planar YUV 4:4:4, 24bpp, full scale (JPEG), deprecated in favor of AV_PIX_FMT_YUV444P and setting col...
Definition: pixfmt.h:87
AVClass
Describe the class of an AVClass context structure.
Definition: log.h:66
NULL
#define NULL
Definition: coverity.c:32
ff_vf_mpdecimate
const AVFilter ff_vf_mpdecimate
Definition: vf_mpdecimate.c:240
OFFSET
#define OFFSET(x)
Definition: vf_mpdecimate.c:56
AV_PIX_FMT_YUVJ420P
@ AV_PIX_FMT_YUVJ420P
planar YUV 4:2:0, 12bpp, full scale (JPEG), deprecated in favor of AV_PIX_FMT_YUV420P and setting col...
Definition: pixfmt.h:85
DecimateContext::ref
AVFrame * ref
reference picture
Definition: vf_mpdecimate.c:52
c
Undefined Behavior In the C some operations are like signed integer dereferencing freed accessing outside allocated Undefined Behavior must not occur in a C it is not safe even if the output of undefined operations is unused The unsafety may seem nit picking but Optimizing compilers have in fact optimized code on the assumption that no undefined Behavior occurs Optimizing code based on wrong assumptions can and has in some cases lead to effects beyond the output of computations The signed integer overflow problem in speed critical code Code which is highly optimized and works with signed integers sometimes has the problem that often the output of the computation does not c
Definition: undefined.txt:32
av_ts2timestr
#define av_ts2timestr(ts, tb)
Convenience macro, the return value should be used only directly in function arguments but never stan...
Definition: timestamp.h:77
NULL_IF_CONFIG_SMALL
#define NULL_IF_CONFIG_SMALL(x)
Return NULL if CONFIG_SMALL is true, otherwise the argument without modification.
Definition: internal.h:106
mpdecimate_options
static const AVOption mpdecimate_options[]
Definition: vf_mpdecimate.c:59
pixelutils.h
init
static av_cold int init(AVFilterContext *ctx)
Definition: vf_mpdecimate.c:151
AV_PIX_FMT_YUVA444P
@ AV_PIX_FMT_YUVA444P
planar YUV 4:4:4 32bpp, (1 Cr & Cb sample per 1x1 Y & A samples)
Definition: pixfmt.h:174
internal.h
AV_OPT_TYPE_FLOAT
@ AV_OPT_TYPE_FLOAT
Definition: opt.h:238
DecimateContext::sad
av_pixelutils_sad_fn sad
sum of absolute difference function
Definition: vf_mpdecimate.c:53
DecimateContext::max_drop_count
int max_drop_count
if positive: maximum number of sequential frames to drop if negative: minimum number of frames betwee...
Definition: vf_mpdecimate.c:42
DecimateContext::frac
float frac
threshold of changed pixels over the total fraction
Definition: vf_mpdecimate.c:40
FFMIN
#define FFMIN(a, b)
Definition: macros.h:49
AV_PIX_FMT_YUVJ440P
@ AV_PIX_FMT_YUVJ440P
planar YUV 4:4:0 full scale (JPEG), deprecated in favor of AV_PIX_FMT_YUV440P and setting color_range
Definition: pixfmt.h:107
AVFilterPad::name
const char * name
Pad name.
Definition: internal.h:39
AVFilter
Filter definition.
Definition: avfilter.h:166
DecimateContext::hi
int hi
lower and higher threshold number of differences values for 8x8 blocks
Definition: vf_mpdecimate.c:37
ret
ret
Definition: filter_design.txt:187
AVFILTER_DEFINE_CLASS
AVFILTER_DEFINE_CLASS(mpdecimate)
DecimateContext::max_keep_count
int max_keep_count
number of similar frames to ignore before to start dropping them
Definition: vf_mpdecimate.c:48
AV_PIX_FMT_NONE
@ AV_PIX_FMT_NONE
Definition: pixfmt.h:72
AV_OPT_TYPE_INT
@ AV_OPT_TYPE_INT
Definition: opt.h:235
avfilter.h
av_pixelutils_get_sad_fn
av_pixelutils_sad_fn av_pixelutils_get_sad_fn(int w_bits, int h_bits, int aligned, void *log_ctx)
Get a potentially optimized pointer to a Sum-of-absolute-differences function (see the av_pixelutils_...
Definition: pixelutils.c:72
ref
static int ref[MAX_W *MAX_W]
Definition: jpeg2000dwt.c:112
AV_PIX_FMT_YUV444P
@ AV_PIX_FMT_YUV444P
planar YUV 4:4:4, 24bpp, (1 Cr & Cb sample per 1x1 Y samples)
Definition: pixfmt.h:78
AVFilterContext
An instance of a filter.
Definition: avfilter.h:407
AV_PIX_FMT_GBRP
@ AV_PIX_FMT_GBRP
planar GBR 4:4:4 24bpp
Definition: pixfmt.h:165
AVMEDIA_TYPE_VIDEO
@ AVMEDIA_TYPE_VIDEO
Definition: avutil.h:201
AV_PIX_FMT_YUV422P
@ AV_PIX_FMT_YUV422P
planar YUV 4:2:2, 16bpp, (1 Cr & Cb sample per 2x1 Y samples)
Definition: pixfmt.h:77
AVPixFmtDescriptor
Descriptor that unambiguously describes how the bits of a pixel are stored in the up to 4 data planes...
Definition: pixdesc.h:69
FILTER_OUTPUTS
#define FILTER_OUTPUTS(array)
Definition: internal.h:183
AV_PIX_FMT_YUV411P
@ AV_PIX_FMT_YUV411P
planar YUV 4:1:1, 12bpp, (1 Cr & Cb sample per 4x1 Y samples)
Definition: pixfmt.h:80
DecimateContext
Definition: vf_decimate.c:38
mpdecimate_inputs
static const AVFilterPad mpdecimate_inputs[]
Definition: vf_mpdecimate.c:231
uninit
static av_cold void uninit(AVFilterContext *ctx)
Definition: vf_mpdecimate.c:165
d
d
Definition: ffmpeg_filter.c:425
timestamp.h
AVFrame::linesize
int linesize[AV_NUM_DATA_POINTERS]
For video, a positive or negative value, which is typically indicating the size in bytes of each pict...
Definition: frame.h:385
AV_PIX_FMT_YUV410P
@ AV_PIX_FMT_YUV410P
planar YUV 4:1:0, 9bpp, (1 Cr & Cb sample per 4x4 Y samples)
Definition: pixfmt.h:79
av_log
#define av_log(a,...)
Definition: tableprint_vlc.h:27
av_ts2str
#define av_ts2str(ts)
Convenience macro, the return value should be used only directly in function arguments but never stan...
Definition: timestamp.h:54
h
h
Definition: vp9dsp_template.c:2038
decimate_frame
static int decimate_frame(AVFilterContext *ctx, AVFrame *cur, AVFrame *ref)
Tell if the frame should be decimated, for example if it is no much different with respect to the ref...
Definition: vf_mpdecimate.c:113
AVPixFmtDescriptor::log2_chroma_h
uint8_t log2_chroma_h
Amount to shift the luma height right to find the chroma height.
Definition: pixdesc.h:89
AV_PIX_FMT_YUVA422P
@ AV_PIX_FMT_YUVA422P
planar YUV 4:2:2 24bpp, (1 Cr & Cb sample per 2x1 Y & A samples)
Definition: pixfmt.h:173
diff_planes
static int diff_planes(AVFilterContext *ctx, uint8_t *cur, int cur_linesize, uint8_t *ref, int ref_linesize, int w, int h)
Return 1 if the two planes are different, 0 otherwise.
Definition: vf_mpdecimate.c:75