1 /*
2 * Copyright (c) 2017 Paul B Mahol
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 "libavutil/imgutils.h"
22 #include "libavutil/opt.h"
23 #include "libavutil/pixdesc.h"
24 #include "libavutil/qsort.h"
25 #include "avfilter.h"
26
27 #define FF_BUFQUEUE_SIZE 129
28 #include "bufferqueue.h"
29
30 #include "formats.h"
31 #include "internal.h"
32 #include "video.h"
33
34 #define SIZE FF_BUFQUEUE_SIZE
35
36 enum smooth_mode {
37 ARITHMETIC_MEAN,
38 GEOMETRIC_MEAN,
39 HARMONIC_MEAN,
40 QUADRATIC_MEAN,
41 CUBIC_MEAN,
42 POWER_MEAN,
43 MEDIAN,
44 NB_SMOOTH_MODE,
45 };
46
47 typedef struct DeflickerContext {
48 const AVClass *class;
49
50 int size;
51 int mode;
52 int bypass;
53
54 int eof;
55 int depth;
56 int nb_planes;
57 int planewidth[4];
58 int planeheight[4];
59
60 uint64_t *histogram;
61 float luminance[SIZE];
62 float sorted[SIZE];
63
64 struct FFBufQueue q;
65 int available;
66
67 void (*get_factor)(AVFilterContext *ctx, float *f);
68 float (*calc_avgy)(AVFilterContext *ctx, AVFrame *in);
69 int (*deflicker)(AVFilterContext *ctx, const uint8_t *src, ptrdiff_t src_linesize,
70 uint8_t *dst, ptrdiff_t dst_linesize, int w, int h, float f);
71 } DeflickerContext;
72
73 #define OFFSET(x) offsetof(DeflickerContext, x)
74 #define FLAGS AV_OPT_FLAG_VIDEO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
75
76 static const AVOption deflicker_options[] = {
77 { "size", "set how many frames to use", OFFSET(size), AV_OPT_TYPE_INT, {.i64=5}, 2, SIZE, FLAGS },
78 { "s", "set how many frames to use", OFFSET(size), AV_OPT_TYPE_INT, {.i64=5}, 2, SIZE, FLAGS },
79 { "mode", "set how to smooth luminance", OFFSET(mode), AV_OPT_TYPE_INT, {.i64=0}, 0, NB_SMOOTH_MODE-1, FLAGS, "mode" },
80 { "m", "set how to smooth luminance", OFFSET(mode), AV_OPT_TYPE_INT, {.i64=0}, 0, NB_SMOOTH_MODE-1, FLAGS, "mode" },
81 { "am", "arithmetic mean", 0, AV_OPT_TYPE_CONST, {.i64=ARITHMETIC_MEAN}, 0, 0, FLAGS, "mode" },
82 { "gm", "geometric mean", 0, AV_OPT_TYPE_CONST, {.i64=GEOMETRIC_MEAN}, 0, 0, FLAGS, "mode" },
83 { "hm", "harmonic mean", 0, AV_OPT_TYPE_CONST, {.i64=HARMONIC_MEAN}, 0, 0, FLAGS, "mode" },
84 { "qm", "quadratic mean", 0, AV_OPT_TYPE_CONST, {.i64=QUADRATIC_MEAN}, 0, 0, FLAGS, "mode" },
85 { "cm", "cubic mean", 0, AV_OPT_TYPE_CONST, {.i64=CUBIC_MEAN}, 0, 0, FLAGS, "mode" },
86 { "pm", "power mean", 0, AV_OPT_TYPE_CONST, {.i64=POWER_MEAN}, 0, 0, FLAGS, "mode" },
87 { "median", "median", 0, AV_OPT_TYPE_CONST, {.i64=MEDIAN}, 0, 0, FLAGS, "mode" },
88 { "bypass", "leave frames unchanged", OFFSET(bypass), AV_OPT_TYPE_BOOL, {.i64=0}, 0, 1, FLAGS },
89 { NULL }
90 };
91
92 AVFILTER_DEFINE_CLASS(deflicker);
93
query_formats(AVFilterContext * ctx)94 static int query_formats(AVFilterContext *ctx)
95 {
96 static const enum AVPixelFormat pixel_fmts[] = {
97 AV_PIX_FMT_GRAY8, AV_PIX_FMT_GRAY9, AV_PIX_FMT_GRAY10,
98 AV_PIX_FMT_GRAY12, AV_PIX_FMT_GRAY14, AV_PIX_FMT_GRAY16,
99 AV_PIX_FMT_YUV410P, AV_PIX_FMT_YUV411P,
100 AV_PIX_FMT_YUV420P, AV_PIX_FMT_YUV422P,
101 AV_PIX_FMT_YUV440P, AV_PIX_FMT_YUV444P,
102 AV_PIX_FMT_YUVJ420P, AV_PIX_FMT_YUVJ422P,
103 AV_PIX_FMT_YUVJ440P, AV_PIX_FMT_YUVJ444P,
104 AV_PIX_FMT_YUVJ411P,
105 AV_PIX_FMT_YUV420P9, AV_PIX_FMT_YUV422P9, AV_PIX_FMT_YUV444P9,
106 AV_PIX_FMT_YUV420P10, AV_PIX_FMT_YUV422P10, AV_PIX_FMT_YUV444P10,
107 AV_PIX_FMT_YUV440P10,
108 AV_PIX_FMT_YUV444P12, AV_PIX_FMT_YUV422P12, AV_PIX_FMT_YUV420P12,
109 AV_PIX_FMT_YUV440P12,
110 AV_PIX_FMT_YUV444P14, AV_PIX_FMT_YUV422P14, AV_PIX_FMT_YUV420P14,
111 AV_PIX_FMT_YUV420P16, AV_PIX_FMT_YUV422P16, AV_PIX_FMT_YUV444P16,
112 AV_PIX_FMT_YUVA420P, AV_PIX_FMT_YUVA422P, AV_PIX_FMT_YUVA444P,
113 AV_PIX_FMT_YUVA444P9, AV_PIX_FMT_YUVA444P10, AV_PIX_FMT_YUVA444P12, AV_PIX_FMT_YUVA444P16,
114 AV_PIX_FMT_YUVA422P9, AV_PIX_FMT_YUVA422P10, AV_PIX_FMT_YUVA422P12, AV_PIX_FMT_YUVA422P16,
115 AV_PIX_FMT_YUVA420P9, AV_PIX_FMT_YUVA420P10, AV_PIX_FMT_YUVA420P16,
116 AV_PIX_FMT_NONE
117 };
118 AVFilterFormats *formats = ff_make_format_list(pixel_fmts);
119 if (!formats)
120 return AVERROR(ENOMEM);
121 return ff_set_common_formats(ctx, formats);
122 }
123
deflicker8(AVFilterContext * ctx,const uint8_t * src,ptrdiff_t src_linesize,uint8_t * dst,ptrdiff_t dst_linesize,int w,int h,float f)124 static int deflicker8(AVFilterContext *ctx,
125 const uint8_t *src, ptrdiff_t src_linesize,
126 uint8_t *dst, ptrdiff_t dst_linesize,
127 int w, int h, float f)
128 {
129 int x, y;
130
131 for (y = 0; y < h; y++) {
132 for (x = 0; x < w; x++) {
133 dst[x] = av_clip_uint8(src[x] * f);
134 }
135
136 dst += dst_linesize;
137 src += src_linesize;
138 }
139
140 return 0;
141 }
142
deflicker16(AVFilterContext * ctx,const uint8_t * ssrc,ptrdiff_t src_linesize,uint8_t * ddst,ptrdiff_t dst_linesize,int w,int h,float f)143 static int deflicker16(AVFilterContext *ctx,
144 const uint8_t *ssrc, ptrdiff_t src_linesize,
145 uint8_t *ddst, ptrdiff_t dst_linesize,
146 int w, int h, float f)
147 {
148 DeflickerContext *s = ctx->priv;
149 const uint16_t *src = (const uint16_t *)ssrc;
150 uint16_t *dst = (uint16_t *)ddst;
151 const int max = (1 << s->depth) - 1;
152 int x, y;
153
154 for (y = 0; y < h; y++) {
155 for (x = 0; x < w; x++) {
156 dst[x] = av_clip(src[x] * f, 0, max);
157 }
158
159 dst += dst_linesize / 2;
160 src += src_linesize / 2;
161 }
162
163 return 0;
164 }
165
calc_avgy8(AVFilterContext * ctx,AVFrame * in)166 static float calc_avgy8(AVFilterContext *ctx, AVFrame *in)
167 {
168 DeflickerContext *s = ctx->priv;
169 const uint8_t *src = in->data[0];
170 int64_t sum = 0;
171 int y, x;
172
173 memset(s->histogram, 0, (1 << s->depth) * sizeof(*s->histogram));
174
175 for (y = 0; y < s->planeheight[0]; y++) {
176 for (x = 0; x < s->planewidth[0]; x++) {
177 s->histogram[src[x]]++;
178 }
179 src += in->linesize[0];
180 }
181
182 for (y = 0; y < 1 << s->depth; y++) {
183 sum += s->histogram[y] * y;
184 }
185
186 return 1.0f * sum / (s->planeheight[0] * s->planewidth[0]);
187 }
188
calc_avgy16(AVFilterContext * ctx,AVFrame * in)189 static float calc_avgy16(AVFilterContext *ctx, AVFrame *in)
190 {
191 DeflickerContext *s = ctx->priv;
192 const uint16_t *src = (const uint16_t *)in->data[0];
193 int64_t sum = 0;
194 int y, x;
195
196 memset(s->histogram, 0, (1 << s->depth) * sizeof(*s->histogram));
197
198 for (y = 0; y < s->planeheight[0]; y++) {
199 for (x = 0; x < s->planewidth[0]; x++) {
200 s->histogram[src[x]]++;
201 }
202 src += in->linesize[0] / 2;
203 }
204
205 for (y = 0; y < 1 << s->depth; y++) {
206 sum += s->histogram[y] * y;
207 }
208
209 return 1.0f * sum / (s->planeheight[0] * s->planewidth[0]);
210 }
211
get_am_factor(AVFilterContext * ctx,float * f)212 static void get_am_factor(AVFilterContext *ctx, float *f)
213 {
214 DeflickerContext *s = ctx->priv;
215 int y;
216
217 *f = 0.0f;
218
219 for (y = 0; y < s->size; y++) {
220 *f += s->luminance[y];
221 }
222
223 *f /= s->size;
224 *f /= s->luminance[0];
225 }
226
get_gm_factor(AVFilterContext * ctx,float * f)227 static void get_gm_factor(AVFilterContext *ctx, float *f)
228 {
229 DeflickerContext *s = ctx->priv;
230 int y;
231
232 *f = 1;
233
234 for (y = 0; y < s->size; y++) {
235 *f *= s->luminance[y];
236 }
237
238 *f = pow(*f, 1.0f / s->size);
239 *f /= s->luminance[0];
240 }
241
get_hm_factor(AVFilterContext * ctx,float * f)242 static void get_hm_factor(AVFilterContext *ctx, float *f)
243 {
244 DeflickerContext *s = ctx->priv;
245 int y;
246
247 *f = 0.0f;
248
249 for (y = 0; y < s->size; y++) {
250 *f += 1.0f / s->luminance[y];
251 }
252
253 *f = s->size / *f;
254 *f /= s->luminance[0];
255 }
256
get_qm_factor(AVFilterContext * ctx,float * f)257 static void get_qm_factor(AVFilterContext *ctx, float *f)
258 {
259 DeflickerContext *s = ctx->priv;
260 int y;
261
262 *f = 0.0f;
263
264 for (y = 0; y < s->size; y++) {
265 *f += s->luminance[y] * s->luminance[y];
266 }
267
268 *f /= s->size;
269 *f = sqrtf(*f);
270 *f /= s->luminance[0];
271 }
272
get_cm_factor(AVFilterContext * ctx,float * f)273 static void get_cm_factor(AVFilterContext *ctx, float *f)
274 {
275 DeflickerContext *s = ctx->priv;
276 int y;
277
278 *f = 0.0f;
279
280 for (y = 0; y < s->size; y++) {
281 *f += s->luminance[y] * s->luminance[y] * s->luminance[y];
282 }
283
284 *f /= s->size;
285 *f = cbrtf(*f);
286 *f /= s->luminance[0];
287 }
288
get_pm_factor(AVFilterContext * ctx,float * f)289 static void get_pm_factor(AVFilterContext *ctx, float *f)
290 {
291 DeflickerContext *s = ctx->priv;
292 int y;
293
294 *f = 0.0f;
295
296 for (y = 0; y < s->size; y++) {
297 *f += powf(s->luminance[y], s->size);
298 }
299
300 *f /= s->size;
301 *f = powf(*f, 1.0f / s->size);
302 *f /= s->luminance[0];
303 }
304
comparef(const void * a,const void * b)305 static int comparef(const void *a, const void *b)
306 {
307 const float *aa = a, *bb = b;
308 return round(aa - bb);
309 }
310
get_median_factor(AVFilterContext * ctx,float * f)311 static void get_median_factor(AVFilterContext *ctx, float *f)
312 {
313 DeflickerContext *s = ctx->priv;
314
315 memcpy(s->sorted, s->luminance, sizeof(s->sorted));
316 AV_QSORT(s->sorted, s->size, float, comparef);
317
318 *f = s->sorted[s->size >> 1] / s->luminance[0];
319 }
320
config_input(AVFilterLink * inlink)321 static int config_input(AVFilterLink *inlink)
322 {
323 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(inlink->format);
324 AVFilterContext *ctx = inlink->dst;
325 DeflickerContext *s = ctx->priv;
326
327 s->nb_planes = desc->nb_components;
328
329 s->planeheight[1] = s->planeheight[2] = AV_CEIL_RSHIFT(inlink->h, desc->log2_chroma_h);
330 s->planeheight[0] = s->planeheight[3] = inlink->h;
331 s->planewidth[1] = s->planewidth[2] = AV_CEIL_RSHIFT(inlink->w, desc->log2_chroma_w);
332 s->planewidth[0] = s->planewidth[3] = inlink->w;
333
334 s->depth = desc->comp[0].depth;
335 if (s->depth == 8) {
336 s->deflicker = deflicker8;
337 s->calc_avgy = calc_avgy8;
338 } else {
339 s->deflicker = deflicker16;
340 s->calc_avgy = calc_avgy16;
341 }
342
343 s->histogram = av_calloc(1 << s->depth, sizeof(*s->histogram));
344 if (!s->histogram)
345 return AVERROR(ENOMEM);
346
347 switch (s->mode) {
348 case MEDIAN: s->get_factor = get_median_factor; break;
349 case ARITHMETIC_MEAN: s->get_factor = get_am_factor; break;
350 case GEOMETRIC_MEAN: s->get_factor = get_gm_factor; break;
351 case HARMONIC_MEAN: s->get_factor = get_hm_factor; break;
352 case QUADRATIC_MEAN: s->get_factor = get_qm_factor; break;
353 case CUBIC_MEAN: s->get_factor = get_cm_factor; break;
354 case POWER_MEAN: s->get_factor = get_pm_factor; break;
355 }
356
357 return 0;
358 }
359
filter_frame(AVFilterLink * inlink,AVFrame * buf)360 static int filter_frame(AVFilterLink *inlink, AVFrame *buf)
361 {
362 AVFilterContext *ctx = inlink->dst;
363 AVFilterLink *outlink = ctx->outputs[0];
364 DeflickerContext *s = ctx->priv;
365 AVDictionary **metadata;
366 AVFrame *out, *in;
367 float f;
368 int y;
369
370 if (s->q.available < s->size && !s->eof) {
371 s->luminance[s->available] = s->calc_avgy(ctx, buf);
372 ff_bufqueue_add(ctx, &s->q, buf);
373 s->available++;
374 return 0;
375 }
376
377 in = ff_bufqueue_peek(&s->q, 0);
378
379 out = ff_get_video_buffer(outlink, outlink->w, outlink->h);
380 if (!out) {
381 av_frame_free(&buf);
382 return AVERROR(ENOMEM);
383 }
384
385 s->get_factor(ctx, &f);
386 if (!s->bypass)
387 s->deflicker(ctx, in->data[0], in->linesize[0], out->data[0], out->linesize[0],
388 outlink->w, outlink->h, f);
389 for (y = 1 - s->bypass; y < s->nb_planes; y++) {
390 av_image_copy_plane(out->data[y], out->linesize[y],
391 in->data[y], in->linesize[y],
392 s->planewidth[y] * (1 + (s->depth > 8)), s->planeheight[y]);
393 }
394
395 av_frame_copy_props(out, in);
396 metadata = &out->metadata;
397 if (metadata) {
398 uint8_t value[128];
399
400 snprintf(value, sizeof(value), "%f", s->luminance[0]);
401 av_dict_set(metadata, "lavfi.deflicker.luminance", value, 0);
402
403 snprintf(value, sizeof(value), "%f", s->luminance[0] * f);
404 av_dict_set(metadata, "lavfi.deflicker.new_luminance", value, 0);
405
406 snprintf(value, sizeof(value), "%f", f - 1.0f);
407 av_dict_set(metadata, "lavfi.deflicker.relative_change", value, 0);
408 }
409
410 in = ff_bufqueue_get(&s->q);
411 av_frame_free(&in);
412 memmove(&s->luminance[0], &s->luminance[1], sizeof(*s->luminance) * (s->size - 1));
413 s->luminance[s->available - 1] = s->calc_avgy(ctx, buf);
414 ff_bufqueue_add(ctx, &s->q, buf);
415
416 return ff_filter_frame(outlink, out);
417 }
418
request_frame(AVFilterLink * outlink)419 static int request_frame(AVFilterLink *outlink)
420 {
421 AVFilterContext *ctx = outlink->src;
422 DeflickerContext *s = ctx->priv;
423 int ret;
424
425 ret = ff_request_frame(ctx->inputs[0]);
426 if (ret == AVERROR_EOF && s->available > 0) {
427 AVFrame *buf = ff_bufqueue_peek(&s->q, s->available - 1);
428 if (!buf)
429 return AVERROR(ENOMEM);
430 buf = av_frame_clone(buf);
431 if (!buf)
432 return AVERROR(ENOMEM);
433
434 s->eof = 1;
435 ret = filter_frame(ctx->inputs[0], buf);
436 s->available--;
437 }
438
439 return ret;
440 }
441
uninit(AVFilterContext * ctx)442 static av_cold void uninit(AVFilterContext *ctx)
443 {
444 DeflickerContext *s = ctx->priv;
445
446 ff_bufqueue_discard_all(&s->q);
447 av_freep(&s->histogram);
448 }
449
450 static const AVFilterPad inputs[] = {
451 {
452 .name = "default",
453 .type = AVMEDIA_TYPE_VIDEO,
454 .filter_frame = filter_frame,
455 .config_props = config_input,
456 },
457 { NULL }
458 };
459
460 static const AVFilterPad outputs[] = {
461 {
462 .name = "default",
463 .type = AVMEDIA_TYPE_VIDEO,
464 .request_frame = request_frame,
465 },
466 { NULL }
467 };
468
469 AVFilter ff_vf_deflicker = {
470 .name = "deflicker",
471 .description = NULL_IF_CONFIG_SMALL("Remove temporal frame luminance variations."),
472 .priv_size = sizeof(DeflickerContext),
473 .priv_class = &deflicker_class,
474 .uninit = uninit,
475 .query_formats = query_formats,
476 .inputs = inputs,
477 .outputs = outputs,
478 };
479