FFmpeg coverage


Directory: ../../../ffmpeg/
File: src/libavfilter/vf_signature.c
Date: 2022-12-09 07:38:14
Exec Total Coverage
Lines: 0 398 0.0%
Functions: 0 13 0.0%
Branches: 0 206 0.0%

Line Branch Exec Source
1 /*
2 * Copyright (c) 2017 Gerion Entrup
3 *
4 * This file is part of FFmpeg.
5 *
6 * FFmpeg is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License as published by
8 * the Free Software Foundation; either version 2 of the License, or
9 * (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
14 * GNU General Public License for more details.
15 *
16 * You should have received a copy of the GNU General Public License along
17 * with FFmpeg; if not, write to the Free Software Foundation, Inc.,
18 * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
19 */
20
21 /**
22 * @file
23 * MPEG-7 video signature calculation and lookup filter
24 * @see http://epubs.surrey.ac.uk/531590/1/MPEG-7%20Video%20Signature%20Author%27s%20Copy.pdf
25 */
26
27 #include "libavcodec/put_bits.h"
28 #include "libavformat/avformat.h"
29 #include "libavutil/opt.h"
30 #include "libavutil/avstring.h"
31 #include "libavutil/file_open.h"
32 #include "avfilter.h"
33 #include "internal.h"
34 #include "signature.h"
35 #include "signature_lookup.c"
36
37 #define OFFSET(x) offsetof(SignatureContext, x)
38 #define FLAGS AV_OPT_FLAG_FILTERING_PARAM | AV_OPT_FLAG_VIDEO_PARAM
39 #define BLOCK_LCM (int64_t) 476985600
40
41 static const AVOption signature_options[] = {
42 { "detectmode", "set the detectmode",
43 OFFSET(mode), AV_OPT_TYPE_INT, {.i64 = MODE_OFF}, 0, NB_LOOKUP_MODE-1, FLAGS, "mode" },
44 { "off", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = MODE_OFF}, 0, 0, .flags = FLAGS, "mode" },
45 { "full", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = MODE_FULL}, 0, 0, .flags = FLAGS, "mode" },
46 { "fast", NULL, 0, AV_OPT_TYPE_CONST, {.i64 = MODE_FAST}, 0, 0, .flags = FLAGS, "mode" },
47 { "nb_inputs", "number of inputs",
48 OFFSET(nb_inputs), AV_OPT_TYPE_INT, {.i64 = 1}, 1, INT_MAX, FLAGS },
49 { "filename", "filename for output files",
50 OFFSET(filename), AV_OPT_TYPE_STRING, {.str = ""}, 0, NB_FORMATS-1, FLAGS },
51 { "format", "set output format",
52 OFFSET(format), AV_OPT_TYPE_INT, {.i64 = FORMAT_BINARY}, 0, 1, FLAGS , "format" },
53 { "binary", 0, 0, AV_OPT_TYPE_CONST, {.i64=FORMAT_BINARY}, 0, 0, FLAGS, "format" },
54 { "xml", 0, 0, AV_OPT_TYPE_CONST, {.i64=FORMAT_XML}, 0, 0, FLAGS, "format" },
55 { "th_d", "threshold to detect one word as similar",
56 OFFSET(thworddist), AV_OPT_TYPE_INT, {.i64 = 9000}, 1, INT_MAX, FLAGS },
57 { "th_dc", "threshold to detect all words as similar",
58 OFFSET(thcomposdist), AV_OPT_TYPE_INT, {.i64 = 60000}, 1, INT_MAX, FLAGS },
59 { "th_xh", "threshold to detect frames as similar",
60 OFFSET(thl1), AV_OPT_TYPE_INT, {.i64 = 116}, 1, INT_MAX, FLAGS },
61 { "th_di", "minimum length of matching sequence in frames",
62 OFFSET(thdi), AV_OPT_TYPE_INT, {.i64 = 0}, 0, INT_MAX, FLAGS },
63 { "th_it", "threshold for relation of good to all frames",
64 OFFSET(thit), AV_OPT_TYPE_DOUBLE, {.dbl = 0.5}, 0.0, 1.0, FLAGS },
65 { NULL }
66 };
67
68 AVFILTER_DEFINE_CLASS(signature);
69
70 /* all formats with a separate gray value */
71 static const enum AVPixelFormat pix_fmts[] = {
72 AV_PIX_FMT_GRAY8,
73 AV_PIX_FMT_YUV410P, AV_PIX_FMT_YUV411P,
74 AV_PIX_FMT_YUV420P, AV_PIX_FMT_YUV422P,
75 AV_PIX_FMT_YUV440P, AV_PIX_FMT_YUV444P,
76 AV_PIX_FMT_YUVJ411P, AV_PIX_FMT_YUVJ420P,
77 AV_PIX_FMT_YUVJ422P, AV_PIX_FMT_YUVJ444P,
78 AV_PIX_FMT_YUVJ440P,
79 AV_PIX_FMT_NV12, AV_PIX_FMT_NV21,
80 AV_PIX_FMT_NONE
81 };
82
83 static int config_input(AVFilterLink *inlink)
84 {
85 AVFilterContext *ctx = inlink->dst;
86 SignatureContext *sic = ctx->priv;
87 StreamContext *sc = &(sic->streamcontexts[FF_INLINK_IDX(inlink)]);
88
89 sc->time_base = inlink->time_base;
90 /* test for overflow */
91 sc->divide = (((uint64_t) inlink->w/32) * (inlink->w/32 + 1) * (inlink->h/32 * inlink->h/32 + 1) > INT64_MAX / (BLOCK_LCM * 255));
92 if (sc->divide) {
93 av_log(ctx, AV_LOG_WARNING, "Input dimension too high for precise calculation, numbers will be rounded.\n");
94 }
95 sc->w = inlink->w;
96 sc->h = inlink->h;
97 return 0;
98 }
99
100 static int get_block_size(const Block *b)
101 {
102 return (b->to.y - b->up.y + 1) * (b->to.x - b->up.x + 1);
103 }
104
105 static uint64_t get_block_sum(StreamContext *sc, uint64_t intpic[32][32], const Block *b)
106 {
107 uint64_t sum = 0;
108
109 int x0, y0, x1, y1;
110
111 x0 = b->up.x;
112 y0 = b->up.y;
113 x1 = b->to.x;
114 y1 = b->to.y;
115
116 if (x0-1 >= 0 && y0-1 >= 0) {
117 sum = intpic[y1][x1] + intpic[y0-1][x0-1] - intpic[y1][x0-1] - intpic[y0-1][x1];
118 } else if (x0-1 >= 0) {
119 sum = intpic[y1][x1] - intpic[y1][x0-1];
120 } else if (y0-1 >= 0) {
121 sum = intpic[y1][x1] - intpic[y0-1][x1];
122 } else {
123 sum = intpic[y1][x1];
124 }
125 return sum;
126 }
127
128 static int cmp(const void *x, const void *y)
129 {
130 const uint64_t *a = x, *b = y;
131 return *a < *b ? -1 : ( *a > *b ? 1 : 0 );
132 }
133
134 /**
135 * sets the bit at position pos to 1 in data
136 */
137 static void set_bit(uint8_t* data, size_t pos)
138 {
139 uint8_t mask = 1 << 7-(pos%8);
140 data[pos/8] |= mask;
141 }
142
143 static int filter_frame(AVFilterLink *inlink, AVFrame *picref)
144 {
145 AVFilterContext *ctx = inlink->dst;
146 SignatureContext *sic = ctx->priv;
147 StreamContext *sc = &(sic->streamcontexts[FF_INLINK_IDX(inlink)]);
148 FineSignature* fs;
149
150 static const uint8_t pot3[5] = { 3*3*3*3, 3*3*3, 3*3, 3, 1 };
151 /* indexes of words : 210,217,219,274,334 44,175,233,270,273 57,70,103,237,269 100,285,295,337,354 101,102,111,275,296
152 s2usw = sorted to unsorted wordvec: 44 is at index 5, 57 at index 10...
153 */
154 static const unsigned int wordvec[25] = {44,57,70,100,101,102,103,111,175,210,217,219,233,237,269,270,273,274,275,285,295,296,334,337,354};
155 static const uint8_t s2usw[25] = { 5,10,11, 15, 20, 21, 12, 22, 6, 0, 1, 2, 7, 13, 14, 8, 9, 3, 23, 16, 17, 24, 4, 18, 19};
156
157 uint8_t wordt2b[5] = { 0, 0, 0, 0, 0 }; /* word ternary to binary */
158 uint64_t intpic[32][32];
159 uint64_t rowcount;
160 uint8_t *p = picref->data[0];
161 int inti, intj;
162 int *intjlut;
163
164 uint64_t conflist[DIFFELEM_SIZE];
165 int f = 0, g = 0, w = 0;
166 int32_t dh1 = 1, dh2 = 1, dw1 = 1, dw2 = 1, a, b;
167 int64_t denom;
168 int i, j, k, ternary;
169 uint64_t blocksum;
170 int blocksize;
171 int64_t th; /* threshold */
172 int64_t sum;
173
174 int64_t precfactor = (sc->divide) ? 65536 : BLOCK_LCM;
175
176 /* initialize fs */
177 if (sc->curfinesig) {
178 fs = av_mallocz(sizeof(FineSignature));
179 if (!fs)
180 return AVERROR(ENOMEM);
181 sc->curfinesig->next = fs;
182 fs->prev = sc->curfinesig;
183 sc->curfinesig = fs;
184 } else {
185 fs = sc->curfinesig = sc->finesiglist;
186 sc->curcoarsesig1->first = fs;
187 }
188
189 fs->pts = picref->pts;
190 fs->index = sc->lastindex++;
191
192 memset(intpic, 0, sizeof(uint64_t)*32*32);
193 intjlut = av_malloc_array(inlink->w, sizeof(int));
194 if (!intjlut)
195 return AVERROR(ENOMEM);
196 for (i = 0; i < inlink->w; i++) {
197 intjlut[i] = (i*32)/inlink->w;
198 }
199
200 for (i = 0; i < inlink->h; i++) {
201 inti = (i*32)/inlink->h;
202 for (j = 0; j < inlink->w; j++) {
203 intj = intjlut[j];
204 intpic[inti][intj] += p[j];
205 }
206 p += picref->linesize[0];
207 }
208 av_freep(&intjlut);
209
210 /* The following calculates a summed area table (intpic) and brings the numbers
211 * in intpic to the same denominator.
212 * So you only have to handle the numinator in the following sections.
213 */
214 dh1 = inlink->h / 32;
215 if (inlink->h % 32)
216 dh2 = dh1 + 1;
217 dw1 = inlink->w / 32;
218 if (inlink->w % 32)
219 dw2 = dw1 + 1;
220 denom = (sc->divide) ? dh1 * (int64_t)dh2 * dw1 * dw2 : 1;
221
222 for (i = 0; i < 32; i++) {
223 rowcount = 0;
224 a = 1;
225 if (dh2 > 1) {
226 a = ((inlink->h*(i+1))%32 == 0) ? (inlink->h*(i+1))/32 - 1 : (inlink->h*(i+1))/32;
227 a -= ((inlink->h*i)%32 == 0) ? (inlink->h*i)/32 - 1 : (inlink->h*i)/32;
228 a = (a == dh1)? dh2 : dh1;
229 }
230 for (j = 0; j < 32; j++) {
231 b = 1;
232 if (dw2 > 1) {
233 b = ((inlink->w*(j+1))%32 == 0) ? (inlink->w*(j+1))/32 - 1 : (inlink->w*(j+1))/32;
234 b -= ((inlink->w*j)%32 == 0) ? (inlink->w*j)/32 - 1 : (inlink->w*j)/32;
235 b = (b == dw1)? dw2 : dw1;
236 }
237 rowcount += intpic[i][j] * a * b * precfactor / denom;
238 if (i > 0) {
239 intpic[i][j] = intpic[i-1][j] + rowcount;
240 } else {
241 intpic[i][j] = rowcount;
242 }
243 }
244 }
245
246 denom = (sc->divide) ? 1 : dh1 * (int64_t)dh2 * dw1 * dw2;
247
248 for (i = 0; i < ELEMENT_COUNT; i++) {
249 const ElemCat* elemcat = elements[i];
250 int64_t* elemsignature;
251 uint64_t* sortsignature;
252
253 elemsignature = av_malloc_array(elemcat->elem_count, sizeof(int64_t));
254 if (!elemsignature)
255 return AVERROR(ENOMEM);
256 sortsignature = av_malloc_array(elemcat->elem_count, sizeof(int64_t));
257 if (!sortsignature) {
258 av_freep(&elemsignature);
259 return AVERROR(ENOMEM);
260 }
261
262 for (j = 0; j < elemcat->elem_count; j++) {
263 blocksum = 0;
264 blocksize = 0;
265 for (k = 0; k < elemcat->left_count; k++) {
266 blocksum += get_block_sum(sc, intpic, &elemcat->blocks[j*elemcat->block_count+k]);
267 blocksize += get_block_size(&elemcat->blocks[j*elemcat->block_count+k]);
268 }
269 sum = blocksum / blocksize;
270 if (elemcat->av_elem) {
271 sum -= 128 * precfactor * denom;
272 } else {
273 blocksum = 0;
274 blocksize = 0;
275 for (; k < elemcat->block_count; k++) {
276 blocksum += get_block_sum(sc, intpic, &elemcat->blocks[j*elemcat->block_count+k]);
277 blocksize += get_block_size(&elemcat->blocks[j*elemcat->block_count+k]);
278 }
279 sum -= blocksum / blocksize;
280 conflist[g++] = FFABS(sum * 8 / (precfactor * denom));
281 }
282
283 elemsignature[j] = sum;
284 sortsignature[j] = FFABS(sum);
285 }
286
287 /* get threshold */
288 qsort(sortsignature, elemcat->elem_count, sizeof(uint64_t), cmp);
289 th = sortsignature[(int) (elemcat->elem_count*0.333)];
290
291 /* ternarize */
292 for (j = 0; j < elemcat->elem_count; j++) {
293 if (elemsignature[j] < -th) {
294 ternary = 0;
295 } else if (elemsignature[j] <= th) {
296 ternary = 1;
297 } else {
298 ternary = 2;
299 }
300 fs->framesig[f/5] += ternary * pot3[f%5];
301
302 if (f == wordvec[w]) {
303 fs->words[s2usw[w]/5] += ternary * pot3[wordt2b[s2usw[w]/5]++];
304 if (w < 24)
305 w++;
306 }
307 f++;
308 }
309 av_freep(&elemsignature);
310 av_freep(&sortsignature);
311 }
312
313 /* confidence */
314 qsort(conflist, DIFFELEM_SIZE, sizeof(uint64_t), cmp);
315 fs->confidence = FFMIN(conflist[DIFFELEM_SIZE/2], 255);
316
317 /* coarsesignature */
318 if (sc->coarsecount == 0) {
319 if (sc->curcoarsesig2) {
320 sc->curcoarsesig1 = av_mallocz(sizeof(CoarseSignature));
321 if (!sc->curcoarsesig1)
322 return AVERROR(ENOMEM);
323 sc->curcoarsesig1->first = fs;
324 sc->curcoarsesig2->next = sc->curcoarsesig1;
325 sc->coarseend = sc->curcoarsesig1;
326 }
327 }
328 if (sc->coarsecount == 45) {
329 sc->midcoarse = 1;
330 sc->curcoarsesig2 = av_mallocz(sizeof(CoarseSignature));
331 if (!sc->curcoarsesig2)
332 return AVERROR(ENOMEM);
333 sc->curcoarsesig2->first = fs;
334 sc->curcoarsesig1->next = sc->curcoarsesig2;
335 sc->coarseend = sc->curcoarsesig2;
336 }
337 for (i = 0; i < 5; i++) {
338 set_bit(sc->curcoarsesig1->data[i], fs->words[i]);
339 }
340 /* assuming the actual frame is the last */
341 sc->curcoarsesig1->last = fs;
342 if (sc->midcoarse) {
343 for (i = 0; i < 5; i++) {
344 set_bit(sc->curcoarsesig2->data[i], fs->words[i]);
345 }
346 sc->curcoarsesig2->last = fs;
347 }
348
349 sc->coarsecount = (sc->coarsecount+1)%90;
350
351 /* debug printing finesignature */
352 if (av_log_get_level() == AV_LOG_DEBUG) {
353 av_log(ctx, AV_LOG_DEBUG, "input %d, confidence: %d\n", FF_INLINK_IDX(inlink), fs->confidence);
354
355 av_log(ctx, AV_LOG_DEBUG, "words:");
356 for (i = 0; i < 5; i++) {
357 av_log(ctx, AV_LOG_DEBUG, " %d:", fs->words[i] );
358 av_log(ctx, AV_LOG_DEBUG, " %d", fs->words[i] / pot3[0] );
359 for (j = 1; j < 5; j++)
360 av_log(ctx, AV_LOG_DEBUG, ",%d", fs->words[i] % pot3[j-1] / pot3[j] );
361 av_log(ctx, AV_LOG_DEBUG, ";");
362 }
363 av_log(ctx, AV_LOG_DEBUG, "\n");
364
365 av_log(ctx, AV_LOG_DEBUG, "framesignature:");
366 for (i = 0; i < SIGELEM_SIZE/5; i++) {
367 av_log(ctx, AV_LOG_DEBUG, " %d", fs->framesig[i] / pot3[0] );
368 for (j = 1; j < 5; j++)
369 av_log(ctx, AV_LOG_DEBUG, ",%d", fs->framesig[i] % pot3[j-1] / pot3[j] );
370 }
371 av_log(ctx, AV_LOG_DEBUG, "\n");
372 }
373
374 if (FF_INLINK_IDX(inlink) == 0)
375 return ff_filter_frame(inlink->dst->outputs[0], picref);
376 return 1;
377 }
378
379 static int xml_export(AVFilterContext *ctx, StreamContext *sc, const char* filename)
380 {
381 FineSignature* fs;
382 CoarseSignature* cs;
383 int i, j;
384 FILE* f;
385 unsigned int pot3[5] = { 3*3*3*3, 3*3*3, 3*3, 3, 1 };
386
387 f = avpriv_fopen_utf8(filename, "w");
388 if (!f) {
389 int err = AVERROR(EINVAL);
390 char buf[128];
391 av_strerror(err, buf, sizeof(buf));
392 av_log(ctx, AV_LOG_ERROR, "cannot open xml file %s: %s\n", filename, buf);
393 return err;
394 }
395
396 /* header */
397 fprintf(f, "<?xml version='1.0' encoding='ASCII' ?>\n");
398 fprintf(f, "<Mpeg7 xmlns=\"urn:mpeg:mpeg7:schema:2001\" xmlns:xsi=\"http://www.w3.org/2001/XMLSchema-instance\" xsi:schemaLocation=\"urn:mpeg:mpeg7:schema:2001 schema/Mpeg7-2001.xsd\">\n");
399 fprintf(f, " <DescriptionUnit xsi:type=\"DescriptorCollectionType\">\n");
400 fprintf(f, " <Descriptor xsi:type=\"VideoSignatureType\">\n");
401 fprintf(f, " <VideoSignatureRegion>\n");
402 fprintf(f, " <VideoSignatureSpatialRegion>\n");
403 fprintf(f, " <Pixel>0 0 </Pixel>\n");
404 fprintf(f, " <Pixel>%d %d </Pixel>\n", sc->w - 1, sc->h - 1);
405 fprintf(f, " </VideoSignatureSpatialRegion>\n");
406 fprintf(f, " <StartFrameOfSpatialRegion>0</StartFrameOfSpatialRegion>\n");
407 /* hoping num is 1, other values are vague */
408 fprintf(f, " <MediaTimeUnit>%d</MediaTimeUnit>\n", sc->time_base.den / sc->time_base.num);
409 fprintf(f, " <MediaTimeOfSpatialRegion>\n");
410 fprintf(f, " <StartMediaTimeOfSpatialRegion>0</StartMediaTimeOfSpatialRegion>\n");
411 fprintf(f, " <EndMediaTimeOfSpatialRegion>%" PRIu64 "</EndMediaTimeOfSpatialRegion>\n", sc->coarseend->last->pts);
412 fprintf(f, " </MediaTimeOfSpatialRegion>\n");
413
414 /* coarsesignatures */
415 for (cs = sc->coarsesiglist; cs; cs = cs->next) {
416 fprintf(f, " <VSVideoSegment>\n");
417 fprintf(f, " <StartFrameOfSegment>%" PRIu32 "</StartFrameOfSegment>\n", cs->first->index);
418 fprintf(f, " <EndFrameOfSegment>%" PRIu32 "</EndFrameOfSegment>\n", cs->last->index);
419 fprintf(f, " <MediaTimeOfSegment>\n");
420 fprintf(f, " <StartMediaTimeOfSegment>%" PRIu64 "</StartMediaTimeOfSegment>\n", cs->first->pts);
421 fprintf(f, " <EndMediaTimeOfSegment>%" PRIu64 "</EndMediaTimeOfSegment>\n", cs->last->pts);
422 fprintf(f, " </MediaTimeOfSegment>\n");
423 for (i = 0; i < 5; i++) {
424 fprintf(f, " <BagOfWords>");
425 for (j = 0; j < 31; j++) {
426 uint8_t n = cs->data[i][j];
427 if (j < 30) {
428 fprintf(f, "%d %d %d %d %d %d %d %d ", (n & 0x80) >> 7,
429 (n & 0x40) >> 6,
430 (n & 0x20) >> 5,
431 (n & 0x10) >> 4,
432 (n & 0x08) >> 3,
433 (n & 0x04) >> 2,
434 (n & 0x02) >> 1,
435 (n & 0x01));
436 } else {
437 /* print only 3 bit in last byte */
438 fprintf(f, "%d %d %d ", (n & 0x80) >> 7,
439 (n & 0x40) >> 6,
440 (n & 0x20) >> 5);
441 }
442 }
443 fprintf(f, "</BagOfWords>\n");
444 }
445 fprintf(f, " </VSVideoSegment>\n");
446 }
447
448 /* finesignatures */
449 for (fs = sc->finesiglist; fs; fs = fs->next) {
450 fprintf(f, " <VideoFrame>\n");
451 fprintf(f, " <MediaTimeOfFrame>%" PRIu64 "</MediaTimeOfFrame>\n", fs->pts);
452 /* confidence */
453 fprintf(f, " <FrameConfidence>%d</FrameConfidence>\n", fs->confidence);
454 /* words */
455 fprintf(f, " <Word>");
456 for (i = 0; i < 5; i++) {
457 fprintf(f, "%d ", fs->words[i]);
458 if (i < 4) {
459 fprintf(f, " ");
460 }
461 }
462 fprintf(f, "</Word>\n");
463 /* framesignature */
464 fprintf(f, " <FrameSignature>");
465 for (i = 0; i< SIGELEM_SIZE/5; i++) {
466 if (i > 0) {
467 fprintf(f, " ");
468 }
469 fprintf(f, "%d ", fs->framesig[i] / pot3[0]);
470 for (j = 1; j < 5; j++)
471 fprintf(f, " %d ", fs->framesig[i] % pot3[j-1] / pot3[j] );
472 }
473 fprintf(f, "</FrameSignature>\n");
474 fprintf(f, " </VideoFrame>\n");
475 }
476 fprintf(f, " </VideoSignatureRegion>\n");
477 fprintf(f, " </Descriptor>\n");
478 fprintf(f, " </DescriptionUnit>\n");
479 fprintf(f, "</Mpeg7>\n");
480
481 fclose(f);
482 return 0;
483 }
484
485 static int binary_export(AVFilterContext *ctx, StreamContext *sc, const char* filename)
486 {
487 FILE* f;
488 FineSignature* fs;
489 CoarseSignature* cs;
490 uint32_t numofsegments = (sc->lastindex + 44)/45;
491 int i, j;
492 PutBitContext buf;
493 /* buffer + header + coarsesignatures + finesignature */
494 int len = (512 + 6 * 32 + 3*16 + 2 +
495 numofsegments * (4*32 + 1 + 5*243) +
496 sc->lastindex * (2 + 32 + 6*8 + 608)) / 8;
497 uint8_t* buffer = av_malloc_array(len, sizeof(uint8_t));
498 if (!buffer)
499 return AVERROR(ENOMEM);
500
501 f = avpriv_fopen_utf8(filename, "wb");
502 if (!f) {
503 int err = AVERROR(EINVAL);
504 char buf[128];
505 av_strerror(err, buf, sizeof(buf));
506 av_log(ctx, AV_LOG_ERROR, "cannot open file %s: %s\n", filename, buf);
507 av_freep(&buffer);
508 return err;
509 }
510 init_put_bits(&buf, buffer, len);
511
512 put_bits32(&buf, 1); /* NumOfSpatial Regions, only 1 supported */
513 put_bits(&buf, 1, 1); /* SpatialLocationFlag, always the whole image */
514 put_bits32(&buf, 0); /* PixelX,1 PixelY,1, 0,0 */
515 put_bits(&buf, 16, sc->w-1 & 0xFFFF); /* PixelX,2 */
516 put_bits(&buf, 16, sc->h-1 & 0xFFFF); /* PixelY,2 */
517 put_bits32(&buf, 0); /* StartFrameOfSpatialRegion */
518 put_bits32(&buf, sc->lastindex); /* NumOfFrames */
519 /* hoping num is 1, other values are vague */
520 /* den/num might be greater than 16 bit, so cutting it */
521 put_bits(&buf, 16, 0xFFFF & (sc->time_base.den / sc->time_base.num)); /* MediaTimeUnit */
522 put_bits(&buf, 1, 1); /* MediaTimeFlagOfSpatialRegion */
523 put_bits32(&buf, 0); /* StartMediaTimeOfSpatialRegion */
524 put_bits32(&buf, 0xFFFFFFFF & sc->coarseend->last->pts); /* EndMediaTimeOfSpatialRegion */
525 put_bits32(&buf, numofsegments); /* NumOfSegments */
526 /* coarsesignatures */
527 for (cs = sc->coarsesiglist; cs; cs = cs->next) {
528 put_bits32(&buf, cs->first->index); /* StartFrameOfSegment */
529 put_bits32(&buf, cs->last->index); /* EndFrameOfSegment */
530 put_bits(&buf, 1, 1); /* MediaTimeFlagOfSegment */
531 put_bits32(&buf, 0xFFFFFFFF & cs->first->pts); /* StartMediaTimeOfSegment */
532 put_bits32(&buf, 0xFFFFFFFF & cs->last->pts); /* EndMediaTimeOfSegment */
533 for (i = 0; i < 5; i++) {
534 /* put 243 bits ( = 7 * 32 + 19 = 8 * 28 + 19) into buffer */
535 for (j = 0; j < 30; j++) {
536 put_bits(&buf, 8, cs->data[i][j]);
537 }
538 put_bits(&buf, 3, cs->data[i][30] >> 5);
539 }
540 }
541 /* finesignatures */
542 put_bits(&buf, 1, 0); /* CompressionFlag, only 0 supported */
543 for (fs = sc->finesiglist; fs; fs = fs->next) {
544 put_bits(&buf, 1, 1); /* MediaTimeFlagOfFrame */
545 put_bits32(&buf, 0xFFFFFFFF & fs->pts); /* MediaTimeOfFrame */
546 put_bits(&buf, 8, fs->confidence); /* FrameConfidence */
547 for (i = 0; i < 5; i++) {
548 put_bits(&buf, 8, fs->words[i]); /* Words */
549 }
550 /* framesignature */
551 for (i = 0; i < SIGELEM_SIZE/5; i++) {
552 put_bits(&buf, 8, fs->framesig[i]);
553 }
554 }
555
556 flush_put_bits(&buf);
557 fwrite(buffer, 1, put_bytes_output(&buf), f);
558 fclose(f);
559 av_freep(&buffer);
560 return 0;
561 }
562
563 static int export(AVFilterContext *ctx, StreamContext *sc, int input)
564 {
565 SignatureContext* sic = ctx->priv;
566 char filename[1024];
567
568 if (sic->nb_inputs > 1) {
569 /* error already handled */
570 av_assert0(av_get_frame_filename(filename, sizeof(filename), sic->filename, input) == 0);
571 } else {
572 if (av_strlcpy(filename, sic->filename, sizeof(filename)) >= sizeof(filename))
573 return AVERROR(EINVAL);
574 }
575 if (sic->format == FORMAT_XML) {
576 return xml_export(ctx, sc, filename);
577 } else {
578 return binary_export(ctx, sc, filename);
579 }
580 }
581
582 static int request_frame(AVFilterLink *outlink)
583 {
584 AVFilterContext *ctx = outlink->src;
585 SignatureContext *sic = ctx->priv;
586 StreamContext *sc, *sc2;
587 MatchingInfo match;
588 int i, j, ret;
589 int lookup = 1; /* indicates wheather EOF of all files is reached */
590
591 /* process all inputs */
592 for (i = 0; i < sic->nb_inputs; i++){
593 sc = &(sic->streamcontexts[i]);
594
595 ret = ff_request_frame(ctx->inputs[i]);
596
597 /* return if unexpected error occurs in input stream */
598 if (ret < 0 && ret != AVERROR_EOF)
599 return ret;
600
601 /* export signature at EOF */
602 if (ret == AVERROR_EOF && !sc->exported) {
603 /* export if wanted */
604 if (strlen(sic->filename) > 0) {
605 if (export(ctx, sc, i) < 0)
606 return ret;
607 }
608 sc->exported = 1;
609 }
610 lookup &= sc->exported;
611 }
612
613 /* signature lookup */
614 if (lookup && sic->mode != MODE_OFF) {
615 /* iterate over every pair */
616 for (i = 0; i < sic->nb_inputs; i++) {
617 sc = &(sic->streamcontexts[i]);
618 for (j = i+1; j < sic->nb_inputs; j++) {
619 sc2 = &(sic->streamcontexts[j]);
620 match = lookup_signatures(ctx, sic, sc, sc2, sic->mode);
621 if (match.score != 0) {
622 av_log(ctx, AV_LOG_INFO, "matching of video %d at %f and %d at %f, %d frames matching\n",
623 i, ((double) match.first->pts * sc->time_base.num) / sc->time_base.den,
624 j, ((double) match.second->pts * sc2->time_base.num) / sc2->time_base.den,
625 match.matchframes);
626 if (match.whole)
627 av_log(ctx, AV_LOG_INFO, "whole video matching\n");
628 } else {
629 av_log(ctx, AV_LOG_INFO, "no matching of video %d and %d\n", i, j);
630 }
631 }
632 }
633 }
634
635 return ret;
636 }
637
638 static av_cold int init(AVFilterContext *ctx)
639 {
640
641 SignatureContext *sic = ctx->priv;
642 StreamContext *sc;
643 int i, ret;
644 char tmp[1024];
645
646 sic->streamcontexts = av_mallocz(sic->nb_inputs * sizeof(StreamContext));
647 if (!sic->streamcontexts)
648 return AVERROR(ENOMEM);
649
650 for (i = 0; i < sic->nb_inputs; i++) {
651 AVFilterPad pad = {
652 .type = AVMEDIA_TYPE_VIDEO,
653 .name = av_asprintf("in%d", i),
654 .config_props = config_input,
655 .filter_frame = filter_frame,
656 };
657
658 if (!pad.name)
659 return AVERROR(ENOMEM);
660 if ((ret = ff_append_inpad_free_name(ctx, &pad)) < 0)
661 return ret;
662
663 sc = &(sic->streamcontexts[i]);
664
665 sc->lastindex = 0;
666 sc->finesiglist = av_mallocz(sizeof(FineSignature));
667 if (!sc->finesiglist)
668 return AVERROR(ENOMEM);
669 sc->curfinesig = NULL;
670
671 sc->coarsesiglist = av_mallocz(sizeof(CoarseSignature));
672 if (!sc->coarsesiglist)
673 return AVERROR(ENOMEM);
674 sc->curcoarsesig1 = sc->coarsesiglist;
675 sc->coarseend = sc->coarsesiglist;
676 sc->coarsecount = 0;
677 sc->midcoarse = 0;
678 }
679
680 /* check filename */
681 if (sic->nb_inputs > 1 && strlen(sic->filename) > 0 && av_get_frame_filename(tmp, sizeof(tmp), sic->filename, 0) == -1) {
682 av_log(ctx, AV_LOG_ERROR, "The filename must contain %%d or %%0nd, if you have more than one input.\n");
683 return AVERROR(EINVAL);
684 }
685
686 return 0;
687 }
688
689
690
691 static av_cold void uninit(AVFilterContext *ctx)
692 {
693 SignatureContext *sic = ctx->priv;
694 StreamContext *sc;
695 void* tmp;
696 FineSignature* finsig;
697 CoarseSignature* cousig;
698 int i;
699
700
701 /* free the lists */
702 if (sic->streamcontexts != NULL) {
703 for (i = 0; i < sic->nb_inputs; i++) {
704 sc = &(sic->streamcontexts[i]);
705 finsig = sc->finesiglist;
706 cousig = sc->coarsesiglist;
707
708 while (finsig) {
709 tmp = finsig;
710 finsig = finsig->next;
711 av_freep(&tmp);
712 }
713 sc->finesiglist = NULL;
714
715 while (cousig) {
716 tmp = cousig;
717 cousig = cousig->next;
718 av_freep(&tmp);
719 }
720 sc->coarsesiglist = NULL;
721 }
722 av_freep(&sic->streamcontexts);
723 }
724 }
725
726 static int config_output(AVFilterLink *outlink)
727 {
728 AVFilterContext *ctx = outlink->src;
729 AVFilterLink *inlink = ctx->inputs[0];
730
731 outlink->time_base = inlink->time_base;
732 outlink->frame_rate = inlink->frame_rate;
733 outlink->sample_aspect_ratio = inlink->sample_aspect_ratio;
734 outlink->w = inlink->w;
735 outlink->h = inlink->h;
736
737 return 0;
738 }
739
740 static const AVFilterPad signature_outputs[] = {
741 {
742 .name = "default",
743 .type = AVMEDIA_TYPE_VIDEO,
744 .request_frame = request_frame,
745 .config_props = config_output,
746 },
747 };
748
749 const AVFilter ff_vf_signature = {
750 .name = "signature",
751 .description = NULL_IF_CONFIG_SMALL("Calculate the MPEG-7 video signature"),
752 .priv_size = sizeof(SignatureContext),
753 .priv_class = &signature_class,
754 .init = init,
755 .uninit = uninit,
756 FILTER_OUTPUTS(signature_outputs),
757 .inputs = NULL,
758 FILTER_PIXFMTS_ARRAY(pix_fmts),
759 .flags = AVFILTER_FLAG_DYNAMIC_INPUTS,
760 };
761