| Line | Branch | Exec | Source |
|---|---|---|---|
| 1 | /* | ||
| 2 | * This file is part of FFmpeg. | ||
| 3 | * | ||
| 4 | * FFmpeg is free software; you can redistribute it and/or | ||
| 5 | * modify it under the terms of the GNU Lesser General Public | ||
| 6 | * License as published by the Free Software Foundation; either | ||
| 7 | * version 2.1 of the License, or (at your option) any later version. | ||
| 8 | * | ||
| 9 | * FFmpeg is distributed in the hope that it will be useful, | ||
| 10 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | ||
| 11 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | ||
| 12 | * Lesser General Public License for more details. | ||
| 13 | * | ||
| 14 | * You should have received a copy of the GNU Lesser General Public | ||
| 15 | * License along with FFmpeg; if not, write to the Free Software | ||
| 16 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA | ||
| 17 | */ | ||
| 18 | |||
| 19 | #include "libavutil/mem.h" | ||
| 20 | #include "libavutil/opt.h" | ||
| 21 | #include "libavutil/pixdesc.h" | ||
| 22 | #include "avfilter.h" | ||
| 23 | #include "filters.h" | ||
| 24 | #include "video.h" | ||
| 25 | |||
| 26 | typedef struct BackgroundkeyContext { | ||
| 27 | const AVClass *class; | ||
| 28 | |||
| 29 | float threshold; | ||
| 30 | float similarity; | ||
| 31 | float blend; | ||
| 32 | int max; | ||
| 33 | |||
| 34 | int nb_threads; | ||
| 35 | int hsub_log2; | ||
| 36 | int vsub_log2; | ||
| 37 | |||
| 38 | int64_t max_sum; | ||
| 39 | int64_t *sums; | ||
| 40 | |||
| 41 | AVFrame *background; | ||
| 42 | |||
| 43 | int (*do_slice)(AVFilterContext *avctx, void *arg, | ||
| 44 | int jobnr, int nb_jobs); | ||
| 45 | } BackgroundkeyContext; | ||
| 46 | |||
| 47 | ✗ | static int do_backgroundkey_slice(AVFilterContext *avctx, void *arg, int jobnr, int nb_jobs) | |
| 48 | { | ||
| 49 | ✗ | BackgroundkeyContext *s = avctx->priv; | |
| 50 | ✗ | AVFrame *frame = arg; | |
| 51 | ✗ | const int slice_start = (frame->height * jobnr) / nb_jobs; | |
| 52 | ✗ | const int slice_end = (frame->height * (jobnr + 1)) / nb_jobs; | |
| 53 | ✗ | const int min_diff = (255 + 255 + 255) * s->similarity; | |
| 54 | ✗ | const float blend = s->blend; | |
| 55 | ✗ | const int hsub = s->hsub_log2; | |
| 56 | ✗ | const int vsub = s->vsub_log2; | |
| 57 | ✗ | int64_t sum = 0; | |
| 58 | |||
| 59 | ✗ | for (int y = slice_start; y < slice_end; y++) { | |
| 60 | ✗ | const uint8_t *srcy = frame->data[0] + frame->linesize[0] * y; | |
| 61 | ✗ | const uint8_t *srcu = frame->data[1] + frame->linesize[1] * (y >> vsub); | |
| 62 | ✗ | const uint8_t *srcv = frame->data[2] + frame->linesize[2] * (y >> vsub); | |
| 63 | ✗ | const uint8_t *bsrcy = s->background->data[0] + s->background->linesize[0] * y; | |
| 64 | ✗ | const uint8_t *bsrcu = s->background->data[1] + s->background->linesize[1] * (y >> vsub); | |
| 65 | ✗ | const uint8_t *bsrcv = s->background->data[2] + s->background->linesize[2] * (y >> vsub); | |
| 66 | ✗ | uint8_t *dst = frame->data[3] + frame->linesize[3] * y; | |
| 67 | ✗ | for (int x = 0; x < frame->width; x++) { | |
| 68 | ✗ | const int xx = x >> hsub; | |
| 69 | ✗ | const int diff = FFABS(srcy[x] - bsrcy[x]) + | |
| 70 | ✗ | FFABS(srcu[xx] - bsrcu[xx]) + | |
| 71 | ✗ | FFABS(srcv[xx] - bsrcv[xx]); | |
| 72 | int A; | ||
| 73 | |||
| 74 | ✗ | sum += diff; | |
| 75 | ✗ | if (blend > 0.f) { | |
| 76 | ✗ | A = 255 - av_clipf((min_diff - diff) / blend, 0.f, 255.f); | |
| 77 | } else { | ||
| 78 | ✗ | A = (diff > min_diff) ? 255 : 0; | |
| 79 | } | ||
| 80 | |||
| 81 | ✗ | dst[x] = A; | |
| 82 | } | ||
| 83 | } | ||
| 84 | |||
| 85 | ✗ | s->sums[jobnr] = sum; | |
| 86 | |||
| 87 | ✗ | return 0; | |
| 88 | } | ||
| 89 | |||
| 90 | ✗ | static int do_backgroundkey16_slice(AVFilterContext *avctx, void *arg, int jobnr, int nb_jobs) | |
| 91 | { | ||
| 92 | ✗ | BackgroundkeyContext *s = avctx->priv; | |
| 93 | ✗ | AVFrame *frame = arg; | |
| 94 | ✗ | const int slice_start = (frame->height * jobnr) / nb_jobs; | |
| 95 | ✗ | const int slice_end = (frame->height * (jobnr + 1)) / nb_jobs; | |
| 96 | ✗ | const int hsub = s->hsub_log2; | |
| 97 | ✗ | const int vsub = s->vsub_log2; | |
| 98 | ✗ | const int max = s->max; | |
| 99 | ✗ | const int min_diff = s->similarity * (s->max + s->max + s->max); | |
| 100 | ✗ | const float blend = s->blend; | |
| 101 | ✗ | int64_t sum = 0; | |
| 102 | |||
| 103 | ✗ | for (int y = slice_start; y < slice_end; y++) { | |
| 104 | ✗ | const uint16_t *srcy = (const uint16_t *)(frame->data[0] + frame->linesize[0] * y); | |
| 105 | ✗ | const uint16_t *srcu = (const uint16_t *)(frame->data[1] + frame->linesize[1] * (y >> vsub)); | |
| 106 | ✗ | const uint16_t *srcv = (const uint16_t *)(frame->data[2] + frame->linesize[2] * (y >> vsub)); | |
| 107 | ✗ | const uint16_t *bsrcy = (const uint16_t *)(s->background->data[0] + s->background->linesize[0] * y); | |
| 108 | ✗ | const uint16_t *bsrcu = (const uint16_t *)(s->background->data[1] + s->background->linesize[1] * (y >> vsub)); | |
| 109 | ✗ | const uint16_t *bsrcv = (const uint16_t *)(s->background->data[2] + s->background->linesize[2] * (y >> vsub)); | |
| 110 | ✗ | uint16_t *dst = (uint16_t *)(frame->data[3] + frame->linesize[3] * y); | |
| 111 | ✗ | for (int x = 0; x < frame->width; x++) { | |
| 112 | ✗ | const int xx = x >> hsub; | |
| 113 | ✗ | const int diff = FFABS(srcy[x] - bsrcy[x] ) + | |
| 114 | ✗ | FFABS(srcu[xx] - bsrcu[xx]) + | |
| 115 | ✗ | FFABS(srcv[xx] - bsrcv[xx]); | |
| 116 | int A; | ||
| 117 | |||
| 118 | ✗ | sum += diff; | |
| 119 | ✗ | if (blend > 0.f) { | |
| 120 | ✗ | A = max - av_clipf((min_diff - diff) / blend, 0.f, max); | |
| 121 | } else { | ||
| 122 | ✗ | A = (diff > min_diff) ? max : 0; | |
| 123 | } | ||
| 124 | |||
| 125 | ✗ | dst[x] = A; | |
| 126 | } | ||
| 127 | } | ||
| 128 | |||
| 129 | ✗ | s->sums[jobnr] = sum; | |
| 130 | |||
| 131 | ✗ | return 0; | |
| 132 | } | ||
| 133 | |||
| 134 | ✗ | static int filter_frame(AVFilterLink *link, AVFrame *frame) | |
| 135 | { | ||
| 136 | ✗ | AVFilterContext *avctx = link->dst; | |
| 137 | ✗ | BackgroundkeyContext *s = avctx->priv; | |
| 138 | ✗ | int64_t sum = 0; | |
| 139 | ✗ | int ret = 0; | |
| 140 | |||
| 141 | ✗ | if (!s->background) { | |
| 142 | ✗ | s->background = ff_get_video_buffer(link, frame->width, frame->height); | |
| 143 | ✗ | if (!s->background) { | |
| 144 | ✗ | ret = AVERROR(ENOMEM); | |
| 145 | ✗ | goto fail; | |
| 146 | } | ||
| 147 | ✗ | ret = av_frame_copy(s->background, frame); | |
| 148 | ✗ | if (ret < 0) | |
| 149 | ✗ | goto fail; | |
| 150 | } | ||
| 151 | |||
| 152 | ✗ | if (ret = ff_filter_execute(avctx, s->do_slice, frame, NULL, | |
| 153 | ✗ | FFMIN(frame->height, s->nb_threads))) | |
| 154 | ✗ | goto fail; | |
| 155 | |||
| 156 | ✗ | for (int n = 0; n < s->nb_threads; n++) | |
| 157 | ✗ | sum += s->sums[n]; | |
| 158 | ✗ | if (s->max_sum * s->threshold < sum) { | |
| 159 | ✗ | ret = av_frame_copy(s->background, frame); | |
| 160 | ✗ | if (ret < 0) | |
| 161 | ✗ | goto fail; | |
| 162 | } | ||
| 163 | |||
| 164 | ✗ | return ff_filter_frame(avctx->outputs[0], frame); | |
| 165 | ✗ | fail: | |
| 166 | ✗ | av_frame_free(&frame); | |
| 167 | ✗ | return ret; | |
| 168 | } | ||
| 169 | |||
| 170 | ✗ | static av_cold int config_output(AVFilterLink *outlink) | |
| 171 | { | ||
| 172 | ✗ | const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(outlink->format); | |
| 173 | ✗ | AVFilterContext *avctx = outlink->src; | |
| 174 | ✗ | AVFilterLink *inlink = avctx->inputs[0]; | |
| 175 | ✗ | BackgroundkeyContext *s = avctx->priv; | |
| 176 | int depth; | ||
| 177 | |||
| 178 | ✗ | s->nb_threads = ff_filter_get_nb_threads(avctx); | |
| 179 | ✗ | depth = desc->comp[0].depth; | |
| 180 | ✗ | s->do_slice = depth <= 8 ? do_backgroundkey_slice : do_backgroundkey16_slice; | |
| 181 | ✗ | s->max = (1 << depth) - 1; | |
| 182 | ✗ | s->hsub_log2 = desc->log2_chroma_w; | |
| 183 | ✗ | s->vsub_log2 = desc->log2_chroma_h; | |
| 184 | ✗ | s->max_sum = (int64_t)(inlink->w) * inlink->h * s->max; | |
| 185 | ✗ | s->max_sum += 2LL * (inlink->w >> s->hsub_log2) * (inlink->h >> s->vsub_log2) * s->max; | |
| 186 | |||
| 187 | ✗ | s->sums = av_calloc(s->nb_threads, sizeof(*s->sums)); | |
| 188 | ✗ | if (!s->sums) | |
| 189 | ✗ | return AVERROR(ENOMEM); | |
| 190 | |||
| 191 | ✗ | return 0; | |
| 192 | } | ||
| 193 | |||
| 194 | ✗ | static av_cold void uninit(AVFilterContext *ctx) | |
| 195 | { | ||
| 196 | ✗ | BackgroundkeyContext *s = ctx->priv; | |
| 197 | |||
| 198 | ✗ | av_frame_free(&s->background); | |
| 199 | ✗ | av_freep(&s->sums); | |
| 200 | ✗ | } | |
| 201 | |||
| 202 | static const AVFilterPad backgroundkey_inputs[] = { | ||
| 203 | { | ||
| 204 | .name = "default", | ||
| 205 | .type = AVMEDIA_TYPE_VIDEO, | ||
| 206 | .flags = AVFILTERPAD_FLAG_NEEDS_WRITABLE, | ||
| 207 | .filter_frame = filter_frame, | ||
| 208 | }, | ||
| 209 | }; | ||
| 210 | |||
| 211 | static const AVFilterPad backgroundkey_outputs[] = { | ||
| 212 | { | ||
| 213 | .name = "default", | ||
| 214 | .type = AVMEDIA_TYPE_VIDEO, | ||
| 215 | .config_props = config_output, | ||
| 216 | }, | ||
| 217 | }; | ||
| 218 | |||
| 219 | #define OFFSET(x) offsetof(BackgroundkeyContext, x) | ||
| 220 | #define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM|AV_OPT_FLAG_RUNTIME_PARAM | ||
| 221 | |||
| 222 | static const AVOption backgroundkey_options[] = { | ||
| 223 | { "threshold", "set the scene change threshold", OFFSET(threshold), AV_OPT_TYPE_FLOAT, { .dbl = 0.08}, 0.0, 1.0, FLAGS }, | ||
| 224 | { "similarity", "set the similarity", OFFSET(similarity), AV_OPT_TYPE_FLOAT, { .dbl = 0.1 }, 0.0, 1.0, FLAGS }, | ||
| 225 | { "blend", "set the blend value", OFFSET(blend), AV_OPT_TYPE_FLOAT, { .dbl = 0.0 }, 0.0, 1.0, FLAGS }, | ||
| 226 | { NULL } | ||
| 227 | }; | ||
| 228 | |||
| 229 | static const enum AVPixelFormat backgroundkey_fmts[] = { | ||
| 230 | AV_PIX_FMT_YUVA420P, | ||
| 231 | AV_PIX_FMT_YUVA422P, | ||
| 232 | AV_PIX_FMT_YUVA444P, | ||
| 233 | AV_PIX_FMT_YUVA420P9, AV_PIX_FMT_YUVA422P9, AV_PIX_FMT_YUVA444P9, | ||
| 234 | AV_PIX_FMT_YUVA420P10, AV_PIX_FMT_YUVA422P10, AV_PIX_FMT_YUVA444P10, | ||
| 235 | AV_PIX_FMT_YUVA422P12, AV_PIX_FMT_YUVA444P12, | ||
| 236 | AV_PIX_FMT_YUVA420P16, AV_PIX_FMT_YUVA422P16, AV_PIX_FMT_YUVA444P16, | ||
| 237 | AV_PIX_FMT_GBRAP, | ||
| 238 | AV_PIX_FMT_GBRAP10, AV_PIX_FMT_GBRAP12, AV_PIX_FMT_GBRAP16, | ||
| 239 | AV_PIX_FMT_NONE | ||
| 240 | }; | ||
| 241 | |||
| 242 | AVFILTER_DEFINE_CLASS(backgroundkey); | ||
| 243 | |||
| 244 | const FFFilter ff_vf_backgroundkey = { | ||
| 245 | .p.name = "backgroundkey", | ||
| 246 | .p.description = NULL_IF_CONFIG_SMALL("Turns a static background into transparency."), | ||
| 247 | .p.priv_class = &backgroundkey_class, | ||
| 248 | .p.flags = AVFILTER_FLAG_SUPPORT_TIMELINE_GENERIC | AVFILTER_FLAG_SLICE_THREADS, | ||
| 249 | .priv_size = sizeof(BackgroundkeyContext), | ||
| 250 | .uninit = uninit, | ||
| 251 | FILTER_INPUTS(backgroundkey_inputs), | ||
| 252 | FILTER_OUTPUTS(backgroundkey_outputs), | ||
| 253 | FILTER_PIXFMTS_ARRAY(backgroundkey_fmts), | ||
| 254 | .process_command = ff_filter_process_command, | ||
| 255 | }; | ||
| 256 |