FFmpeg
vf_photosensitivity.c
Go to the documentation of this file.
1 /*
2  * Copyright (c) 2019 Vladimir Panteleev
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 #include <float.h>
22 
23 #include "libavutil/imgutils.h"
24 #include "libavutil/opt.h"
25 #include "libavutil/pixdesc.h"
26 #include "avfilter.h"
27 
28 #include "formats.h"
29 #include "internal.h"
30 #include "video.h"
31 
32 #define MAX_FRAMES 240
33 #define GRID_SIZE 8
34 #define NUM_CHANNELS 3
35 
36 typedef struct PhotosensitivityFrame {
37  uint8_t grid[GRID_SIZE][GRID_SIZE][4];
39 
40 typedef struct PhotosensitivityContext {
41  const AVClass *class;
42 
43  int nb_frames;
44  int skip;
46  int bypass;
47 
49 
50  /* Circular buffer */
53 
57 
58 #define OFFSET(x) offsetof(PhotosensitivityContext, x)
59 #define FLAGS AV_OPT_FLAG_VIDEO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
60 
62  { "frames", "set how many frames to use", OFFSET(nb_frames), AV_OPT_TYPE_INT, {.i64=30}, 2, MAX_FRAMES, FLAGS },
63  { "f", "set how many frames to use", OFFSET(nb_frames), AV_OPT_TYPE_INT, {.i64=30}, 2, MAX_FRAMES, FLAGS },
64  { "threshold", "set detection threshold factor (lower is stricter)", OFFSET(threshold_multiplier), AV_OPT_TYPE_FLOAT, {.dbl=1}, 0.1, FLT_MAX, FLAGS },
65  { "t", "set detection threshold factor (lower is stricter)", OFFSET(threshold_multiplier), AV_OPT_TYPE_FLOAT, {.dbl=1}, 0.1, FLT_MAX, FLAGS },
66  { "skip", "set pixels to skip when sampling frames", OFFSET(skip), AV_OPT_TYPE_INT, {.i64=1}, 1, 1024, FLAGS },
67  { "bypass", "leave frames unchanged", OFFSET(bypass), AV_OPT_TYPE_BOOL, {.i64=0}, 0, 1, FLAGS },
68  { NULL }
69 };
70 
71 AVFILTER_DEFINE_CLASS(photosensitivity);
72 
74 {
77  int skip;
79 
80 #define NUM_CELLS (GRID_SIZE * GRID_SIZE)
81 
82 static int convert_frame_partial(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
83 {
84  int cell, gx, gy, x0, x1, y0, y1, x, y, c, area;
85  int sum[NUM_CHANNELS];
86  const uint8_t *p;
87 
89 
90  const int slice_start = (NUM_CELLS * jobnr) / nb_jobs;
91  const int slice_end = (NUM_CELLS * (jobnr+1)) / nb_jobs;
92 
93  int width = td->in->width, height = td->in->height, linesize = td->in->linesize[0], skip = td->skip;
94  const uint8_t *data = td->in->data[0];
95 
96  for (cell = slice_start; cell < slice_end; cell++) {
97  gx = cell % GRID_SIZE;
98  gy = cell / GRID_SIZE;
99 
100  x0 = width * gx / GRID_SIZE;
101  x1 = width * (gx+1) / GRID_SIZE;
102  y0 = height * gy / GRID_SIZE;
103  y1 = height * (gy+1) / GRID_SIZE;
104 
105  for (c = 0; c < NUM_CHANNELS; c++) {
106  sum[c] = 0;
107  }
108  for (y = y0; y < y1; y += skip) {
109  p = data + y * linesize + x0 * NUM_CHANNELS;
110  for (x = x0; x < x1; x += skip) {
111  //av_log(NULL, AV_LOG_VERBOSE, "%d %d %d : (%d,%d) (%d,%d) -> %d,%d | *%d\n", c, gx, gy, x0, y0, x1, y1, x, y, (int)row);
112  sum[0] += p[0];
113  sum[1] += p[1];
114  sum[2] += p[2];
115  p += NUM_CHANNELS * skip;
116  // TODO: variable size
117  }
118  }
119 
120  area = ((x1 - x0 + skip - 1) / skip) * ((y1 - y0 + skip - 1) / skip);
121  for (c = 0; c < NUM_CHANNELS; c++) {
122  if (area)
123  sum[c] /= area;
124  td->out->grid[gy][gx][c] = sum[c];
125  }
126  }
127  return 0;
128 }
129 
131 {
133  td.in = in;
134  td.out = out;
135  td.skip = skip;
138 }
139 
141 {
144  uint16_t s_mul;
146 
147 static int blend_frame_partial(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
148 {
149  int x, y;
150  uint8_t *t, *s;
151 
153  const uint16_t s_mul = td->s_mul;
154  const uint16_t t_mul = 0x100 - s_mul;
155  const int slice_start = (td->target->height * jobnr) / nb_jobs;
156  const int slice_end = (td->target->height * (jobnr+1)) / nb_jobs;
157  const int linesize = td->target->linesize[0];
158 
159  for (y = slice_start; y < slice_end; y++) {
160  t = td->target->data[0] + y * td->target->linesize[0];
161  s = td->source->data[0] + y * td->source->linesize[0];
162  for (x = 0; x < linesize; x++) {
163  *t = (*t * t_mul + *s * s_mul) >> 8;
164  t++; s++;
165  }
166  }
167  return 0;
168 }
169 
170 static void blend_frame(AVFilterContext *ctx, AVFrame *target, AVFrame *source, float factor)
171 {
173  td.target = target;
174  td.source = source;
175  td.s_mul = (uint16_t)(factor * 0x100);
177  FFMIN(ctx->outputs[0]->h, ff_filter_get_nb_threads(ctx)));
178 }
179 
181 {
182  int badness, x, y, c;
183  badness = 0;
184  for (c = 0; c < NUM_CHANNELS; c++) {
185  for (y = 0; y < GRID_SIZE; y++) {
186  for (x = 0; x < GRID_SIZE; x++) {
187  badness += abs((int)a->grid[y][x][c] - (int)b->grid[y][x][c]);
188  //av_log(NULL, AV_LOG_VERBOSE, "%d - %d -> %d \n", a->grid[y][x], b->grid[y][x], badness);
189  //av_log(NULL, AV_LOG_VERBOSE, "%d -> %d \n", abs((int)a->grid[y][x] - (int)b->grid[y][x]), badness);
190  }
191  }
192  }
193  return badness;
194 }
195 
197 {
198  /* const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(inlink->format); */
199  AVFilterContext *ctx = inlink->dst;
200  PhotosensitivityContext *s = ctx->priv;
201 
202  s->badness_threshold = (int)(GRID_SIZE * GRID_SIZE * 4 * 256 * s->nb_frames * s->threshold_multiplier / 128);
203 
204  return 0;
205 }
206 
208 {
209  int this_badness, current_badness, fixed_badness, new_badness, i, res;
211  AVFrame *src, *out;
212  int free_in = 0;
213  float factor;
214  AVDictionary **metadata;
215 
216  AVFilterContext *ctx = inlink->dst;
217  AVFilterLink *outlink = ctx->outputs[0];
218  PhotosensitivityContext *s = ctx->priv;
219 
220  /* weighted moving average */
221  current_badness = 0;
222  for (i = 1; i < s->nb_frames; i++)
223  current_badness += i * s->history[(s->history_pos + i) % s->nb_frames];
224  current_badness /= s->nb_frames;
225 
226  convert_frame(ctx, in, &ef, s->skip);
227  this_badness = get_badness(&ef, &s->last_frame_e);
228  new_badness = current_badness + this_badness;
229  av_log(s, AV_LOG_VERBOSE, "badness: %6d -> %6d / %6d (%3d%% - %s)\n",
230  current_badness, new_badness, s->badness_threshold,
231  100 * new_badness / s->badness_threshold, new_badness < s->badness_threshold ? "OK" : "EXCEEDED");
232 
233  fixed_badness = new_badness;
234  if (new_badness < s->badness_threshold || !s->last_frame_av || s->bypass) {
235  factor = 1; /* for metadata */
236  av_frame_free(&s->last_frame_av);
237  s->last_frame_av = src = in;
238  s->last_frame_e = ef;
239  s->history[s->history_pos] = this_badness;
240  } else {
241  factor = (float)(s->badness_threshold - current_badness) / (new_badness - current_badness);
242  if (factor <= 0) {
243  /* just duplicate the frame */
244  s->history[s->history_pos] = 0; /* frame was duplicated, thus, delta is zero */
245  } else {
246  res = av_frame_make_writable(s->last_frame_av);
247  if (res) {
248  av_frame_free(&in);
249  return res;
250  }
251  blend_frame(ctx, s->last_frame_av, in, factor);
252 
253  convert_frame(ctx, s->last_frame_av, &ef, s->skip);
254  this_badness = get_badness(&ef, &s->last_frame_e);
255  fixed_badness = current_badness + this_badness;
256  av_log(s, AV_LOG_VERBOSE, " fixed: %6d -> %6d / %6d (%3d%%) factor=%5.3f\n",
257  current_badness, fixed_badness, s->badness_threshold,
258  100 * new_badness / s->badness_threshold, factor);
259  s->last_frame_e = ef;
260  s->history[s->history_pos] = this_badness;
261  }
262  src = s->last_frame_av;
263  free_in = 1;
264  }
265  s->history_pos = (s->history_pos + 1) % s->nb_frames;
266 
267  out = ff_get_video_buffer(outlink, in->width, in->height);
268  if (!out) {
269  if (free_in == 1)
270  av_frame_free(&in);
271  return AVERROR(ENOMEM);
272  }
274  metadata = &out->metadata;
275  if (metadata) {
276  char value[128];
277 
278  snprintf(value, sizeof(value), "%f", (float)new_badness / s->badness_threshold);
279  av_dict_set(metadata, "lavfi.photosensitivity.badness", value, 0);
280 
281  snprintf(value, sizeof(value), "%f", (float)fixed_badness / s->badness_threshold);
282  av_dict_set(metadata, "lavfi.photosensitivity.fixed-badness", value, 0);
283 
284  snprintf(value, sizeof(value), "%f", (float)this_badness / s->badness_threshold);
285  av_dict_set(metadata, "lavfi.photosensitivity.frame-badness", value, 0);
286 
287  snprintf(value, sizeof(value), "%f", factor);
288  av_dict_set(metadata, "lavfi.photosensitivity.factor", value, 0);
289  }
291  if (free_in == 1)
292  av_frame_free(&in);
293  return ff_filter_frame(outlink, out);
294 }
295 
297 {
298  PhotosensitivityContext *s = ctx->priv;
299 
300  av_frame_free(&s->last_frame_av);
301 }
302 
303 static const AVFilterPad inputs[] = {
304  {
305  .name = "default",
306  .type = AVMEDIA_TYPE_VIDEO,
307  .filter_frame = filter_frame,
308  .config_props = config_input,
309  },
310 };
311 
312 static const AVFilterPad outputs[] = {
313  {
314  .name = "default",
315  .type = AVMEDIA_TYPE_VIDEO,
316  },
317 };
318 
320  .name = "photosensitivity",
321  .description = NULL_IF_CONFIG_SMALL("Filter out photosensitive epilepsy seizure-inducing flashes."),
322  .priv_size = sizeof(PhotosensitivityContext),
323  .priv_class = &photosensitivity_class,
324  .uninit = uninit,
328 };
ff_get_video_buffer
AVFrame * ff_get_video_buffer(AVFilterLink *link, int w, int h)
Request a picture buffer with a specific set of permissions.
Definition: video.c:101
td
#define td
Definition: regdef.h:70
AVFILTER_DEFINE_CLASS
AVFILTER_DEFINE_CLASS(photosensitivity)
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
config_input
static int config_input(AVFilterLink *inlink)
Definition: vf_photosensitivity.c:196
out
FILE * out
Definition: movenc.c:54
ff_filter_frame
int ff_filter_frame(AVFilterLink *link, AVFrame *frame)
Send a frame of data to the next filter.
Definition: avfilter.c:999
PhotosensitivityContext
Definition: vf_photosensitivity.c:40
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
PhotosensitivityContext::history
int history[MAX_FRAMES]
Definition: vf_photosensitivity.c:51
PhotosensitivityContext::badness_threshold
int badness_threshold
Definition: vf_photosensitivity.c:48
av_frame_free
void av_frame_free(AVFrame **frame)
Free the frame and any dynamically allocated objects in it, e.g.
Definition: frame.c:111
AVFrame
This structure describes decoded (raw) audio or video data.
Definition: frame.h:325
av_frame_make_writable
int av_frame_make_writable(AVFrame *frame)
Ensure that the frame data is writable, avoiding data copy if possible.
Definition: frame.c:540
pixdesc.h
AVFrame::width
int width
Definition: frame.h:397
GRID_SIZE
#define GRID_SIZE
Definition: vf_photosensitivity.c:33
AVOption
AVOption.
Definition: opt.h:251
b
#define b
Definition: input.c:34
data
const char data[16]
Definition: mxf.c:143
AV_LOG_VERBOSE
#define AV_LOG_VERBOSE
Detailed information.
Definition: log.h:196
float.h
AV_PIX_FMT_BGR24
@ AV_PIX_FMT_BGR24
packed RGB 8:8:8, 24bpp, BGRBGR...
Definition: pixfmt.h:69
AVDictionary
Definition: dict.c:30
AVFilter::name
const char * name
Filter name.
Definition: avfilter.h:175
video.h
convert_frame_partial
static int convert_frame_partial(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
Definition: vf_photosensitivity.c:82
formats.h
ThreadData_blend_frame::target
AVFrame * target
Definition: vf_photosensitivity.c:142
uninit
static av_cold void uninit(AVFilterContext *ctx)
Definition: vf_photosensitivity.c:296
photosensitivity_options
static const AVOption photosensitivity_options[]
Definition: vf_photosensitivity.c:61
PhotosensitivityContext::last_frame_e
PhotosensitivityFrame last_frame_e
Definition: vf_photosensitivity.c:54
PhotosensitivityContext::last_frame_av
AVFrame * last_frame_av
Definition: vf_photosensitivity.c:55
ThreadData_convert_frame::skip
int skip
Definition: vf_photosensitivity.c:77
ff_vf_photosensitivity
const AVFilter ff_vf_photosensitivity
Definition: vf_photosensitivity.c:319
AVFilterPad
A filter pad used for either input or output.
Definition: internal.h:49
av_cold
#define av_cold
Definition: attributes.h:90
float
float
Definition: af_crystalizer.c:122
width
#define width
s
#define s(width, name)
Definition: cbs_vp9.c:256
ThreadData_convert_frame
Definition: vf_photosensitivity.c:73
slice_end
static int slice_end(AVCodecContext *avctx, AVFrame *pict)
Handle slice ends.
Definition: mpeg12dec.c:2013
PhotosensitivityFrame::grid
uint8_t grid[GRID_SIZE][GRID_SIZE][4]
Definition: vf_photosensitivity.c:37
OFFSET
#define OFFSET(x)
Definition: vf_photosensitivity.c:58
outputs
static const AVFilterPad outputs[]
Definition: vf_photosensitivity.c:312
ctx
AVFormatContext * ctx
Definition: movenc.c:48
FILTER_INPUTS
#define FILTER_INPUTS(array)
Definition: internal.h:190
convert_frame
static void convert_frame(AVFilterContext *ctx, AVFrame *in, PhotosensitivityFrame *out, int skip)
Definition: vf_photosensitivity.c:130
arg
const char * arg
Definition: jacosubdec.c:67
AVClass
Describe the class of an AVClass context structure.
Definition: log.h:66
NULL
#define NULL
Definition: coverity.c:32
av_frame_copy_props
int av_frame_copy_props(AVFrame *dst, const AVFrame *src)
Copy only "metadata" fields from src to dst.
Definition: frame.c:596
FLAGS
#define FLAGS
Definition: vf_photosensitivity.c:59
ThreadData_convert_frame::out
PhotosensitivityFrame * out
Definition: vf_photosensitivity.c:76
PhotosensitivityFrame
Definition: vf_photosensitivity.c:36
abs
#define abs(x)
Definition: cuda_runtime.h:35
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
PhotosensitivityContext::history_pos
int history_pos
Definition: vf_photosensitivity.c:52
source
these buffered frames must be flushed immediately if a new input produces new the filter must not call request_frame to get more It must just process the frame or queue it The task of requesting more frames is left to the filter s request_frame method or the application If a filter has several the filter must be ready for frames arriving randomly on any input any filter with several inputs will most likely require some kind of queuing mechanism It is perfectly acceptable to have a limited queue and to drop frames when the inputs are too unbalanced request_frame For filters that do not use the this method is called when a frame is wanted on an output For a source
Definition: filter_design.txt:255
AV_PIX_FMT_RGB24
@ AV_PIX_FMT_RGB24
packed RGB 8:8:8, 24bpp, RGBRGB...
Definition: pixfmt.h:68
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:117
FILTER_PIXFMTS
#define FILTER_PIXFMTS(...)
Definition: internal.h:176
av_frame_copy
int av_frame_copy(AVFrame *dst, const AVFrame *src)
Copy the frame data from src to dst.
Definition: frame.c:764
filter_frame
static int filter_frame(AVFilterLink *inlink, AVFrame *in)
Definition: vf_photosensitivity.c:207
blend_frame
static void blend_frame(AVFilterContext *ctx, AVFrame *target, AVFrame *source, float factor)
Definition: vf_photosensitivity.c:170
height
#define height
a
The reader does not expect b to be semantically here and if the code is changed by maybe adding a a division or other the signedness will almost certainly be mistaken To avoid this confusion a new type was SUINT is the C unsigned type but it holds a signed int to use the same example SUINT a
Definition: undefined.txt:41
internal.h
AV_OPT_TYPE_FLOAT
@ AV_OPT_TYPE_FLOAT
Definition: opt.h:228
PhotosensitivityContext::skip
int skip
Definition: vf_photosensitivity.c:44
NUM_CHANNELS
#define NUM_CHANNELS
Definition: vf_photosensitivity.c:34
i
#define i(width, name, range_min, range_max)
Definition: cbs_h2645.c:271
ff_filter_get_nb_threads
int ff_filter_get_nb_threads(AVFilterContext *ctx)
Get number of threads for current filter instance.
Definition: avfilter.c:783
value
it s the only field you need to keep assuming you have a context There is some magic you don t need to care about around this just let it vf default value
Definition: writing_filters.txt:86
FFMIN
#define FFMIN(a, b)
Definition: macros.h:49
ThreadData_blend_frame::s_mul
uint16_t s_mul
Definition: vf_photosensitivity.c:144
AVFilterPad::name
const char * name
Pad name.
Definition: internal.h:55
PhotosensitivityContext::bypass
int bypass
Definition: vf_photosensitivity.c:46
AVFilter
Filter definition.
Definition: avfilter.h:171
ThreadData_blend_frame
Definition: vf_photosensitivity.c:140
ThreadData_blend_frame::source
AVFrame * source
Definition: vf_photosensitivity.c:143
MAX_FRAMES
#define MAX_FRAMES
Definition: vf_photosensitivity.c:32
AVFrame::height
int height
Definition: frame.h:397
blend_frame_partial
static int blend_frame_partial(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
Definition: vf_photosensitivity.c:147
AV_OPT_TYPE_INT
@ AV_OPT_TYPE_INT
Definition: opt.h:225
avfilter.h
AVFilterContext
An instance of a filter.
Definition: avfilter.h:408
factor
static const int factor[16]
Definition: vf_pp7.c:76
AVMEDIA_TYPE_VIDEO
@ AVMEDIA_TYPE_VIDEO
Definition: avutil.h:201
inputs
static const AVFilterPad inputs[]
Definition: vf_photosensitivity.c:303
AV_OPT_TYPE_BOOL
@ AV_OPT_TYPE_BOOL
Definition: opt.h:244
FILTER_OUTPUTS
#define FILTER_OUTPUTS(array)
Definition: internal.h:191
av_dict_set
int av_dict_set(AVDictionary **pm, const char *key, const char *value, int flags)
Set the given entry in *pm, overwriting an existing entry.
Definition: dict.c:70
src
INIT_CLIP pixel * src
Definition: h264pred_template.c:418
imgutils.h
av_log
#define av_log(a,...)
Definition: tableprint_vlc.h:27
PhotosensitivityContext::nb_frames
int nb_frames
Definition: vf_photosensitivity.c:43
get_badness
static int get_badness(PhotosensitivityFrame *a, PhotosensitivityFrame *b)
Definition: vf_photosensitivity.c:180
ff_filter_execute
static av_always_inline int ff_filter_execute(AVFilterContext *ctx, avfilter_action_func *func, void *arg, int *ret, int nb_jobs)
Definition: internal.h:142
NUM_CELLS
#define NUM_CELLS
Definition: vf_photosensitivity.c:80
int
int
Definition: ffmpeg_filter.c:153
snprintf
#define snprintf
Definition: snprintf.h:34
PhotosensitivityContext::threshold_multiplier
float threshold_multiplier
Definition: vf_photosensitivity.c:45
ThreadData_convert_frame::in
AVFrame * in
Definition: vf_photosensitivity.c:75