Commit | Line | Data |
---|---|---|
2ba45a60 DM |
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 "formats.h" | |
34 | #include "video.h" | |
35 | ||
36 | typedef struct { | |
37 | const AVClass *class; | |
38 | int lo, hi; ///< lower and higher threshold number of differences | |
39 | ///< values for 8x8 blocks | |
40 | ||
41 | float frac; ///< threshold of changed pixels over the total fraction | |
42 | ||
43 | int max_drop_count; ///< if positive: maximum number of sequential frames to drop | |
44 | ///< if negative: minimum number of frames between two drops | |
45 | ||
46 | int drop_count; ///< if positive: number of frames sequentially dropped | |
47 | ///< if negative: number of sequential frames which were not dropped | |
48 | ||
49 | int hsub, vsub; ///< chroma subsampling values | |
50 | AVFrame *ref; ///< reference picture | |
51 | av_pixelutils_sad_fn sad; ///< sum of absolute difference function | |
52 | } DecimateContext; | |
53 | ||
54 | #define OFFSET(x) offsetof(DecimateContext, x) | |
55 | #define FLAGS AV_OPT_FLAG_VIDEO_PARAM|AV_OPT_FLAG_FILTERING_PARAM | |
56 | ||
57 | static const AVOption mpdecimate_options[] = { | |
58 | { "max", "set the maximum number of consecutive dropped frames (positive), or the minimum interval between dropped frames (negative)", | |
59 | OFFSET(max_drop_count), AV_OPT_TYPE_INT, {.i64=0}, INT_MIN, INT_MAX, FLAGS }, | |
60 | { "hi", "set high dropping threshold", OFFSET(hi), AV_OPT_TYPE_INT, {.i64=64*12}, INT_MIN, INT_MAX, FLAGS }, | |
61 | { "lo", "set low dropping threshold", OFFSET(lo), AV_OPT_TYPE_INT, {.i64=64*5}, INT_MIN, INT_MAX, FLAGS }, | |
62 | { "frac", "set fraction dropping threshold", OFFSET(frac), AV_OPT_TYPE_FLOAT, {.dbl=0.33}, 0, 1, FLAGS }, | |
63 | { NULL } | |
64 | }; | |
65 | ||
66 | AVFILTER_DEFINE_CLASS(mpdecimate); | |
67 | ||
68 | /** | |
69 | * Return 1 if the two planes are different, 0 otherwise. | |
70 | */ | |
71 | static int diff_planes(AVFilterContext *ctx, | |
72 | uint8_t *cur, int cur_linesize, | |
73 | uint8_t *ref, int ref_linesize, | |
74 | int w, int h) | |
75 | { | |
76 | DecimateContext *decimate = ctx->priv; | |
77 | ||
78 | int x, y; | |
79 | int d, c = 0; | |
80 | int t = (w/16)*(h/16)*decimate->frac; | |
81 | ||
82 | /* compute difference for blocks of 8x8 bytes */ | |
83 | for (y = 0; y < h-7; y += 4) { | |
84 | for (x = 8; x < w-7; x += 4) { | |
85 | d = decimate->sad(cur + y*cur_linesize + x, cur_linesize, | |
86 | ref + y*ref_linesize + x, ref_linesize); | |
87 | if (d > decimate->hi) | |
88 | return 1; | |
89 | if (d > decimate->lo) { | |
90 | c++; | |
91 | if (c > t) | |
92 | return 1; | |
93 | } | |
94 | } | |
95 | } | |
96 | return 0; | |
97 | } | |
98 | ||
99 | /** | |
100 | * Tell if the frame should be decimated, for example if it is no much | |
101 | * different with respect to the reference frame ref. | |
102 | */ | |
103 | static int decimate_frame(AVFilterContext *ctx, | |
104 | AVFrame *cur, AVFrame *ref) | |
105 | { | |
106 | DecimateContext *decimate = ctx->priv; | |
107 | int plane; | |
108 | ||
109 | if (decimate->max_drop_count > 0 && | |
110 | decimate->drop_count >= decimate->max_drop_count) | |
111 | return 0; | |
112 | if (decimate->max_drop_count < 0 && | |
113 | (decimate->drop_count-1) > decimate->max_drop_count) | |
114 | return 0; | |
115 | ||
116 | for (plane = 0; ref->data[plane] && ref->linesize[plane]; plane++) { | |
117 | int vsub = plane == 1 || plane == 2 ? decimate->vsub : 0; | |
118 | int hsub = plane == 1 || plane == 2 ? decimate->hsub : 0; | |
119 | if (diff_planes(ctx, | |
120 | cur->data[plane], cur->linesize[plane], | |
121 | ref->data[plane], ref->linesize[plane], | |
122 | FF_CEIL_RSHIFT(ref->width, hsub), | |
123 | FF_CEIL_RSHIFT(ref->height, vsub))) | |
124 | return 0; | |
125 | } | |
126 | ||
127 | return 1; | |
128 | } | |
129 | ||
130 | static av_cold int init(AVFilterContext *ctx) | |
131 | { | |
132 | DecimateContext *decimate = ctx->priv; | |
133 | ||
134 | decimate->sad = av_pixelutils_get_sad_fn(3, 3, 0, decimate); // 8x8, not aligned on blocksize | |
135 | if (!decimate->sad) | |
136 | return AVERROR(EINVAL); | |
137 | ||
138 | av_log(ctx, AV_LOG_VERBOSE, "max_drop_count:%d hi:%d lo:%d frac:%f\n", | |
139 | decimate->max_drop_count, decimate->hi, decimate->lo, decimate->frac); | |
140 | ||
141 | return 0; | |
142 | } | |
143 | ||
144 | static av_cold void uninit(AVFilterContext *ctx) | |
145 | { | |
146 | DecimateContext *decimate = ctx->priv; | |
147 | av_frame_free(&decimate->ref); | |
148 | } | |
149 | ||
150 | static int query_formats(AVFilterContext *ctx) | |
151 | { | |
152 | static const enum AVPixelFormat pix_fmts[] = { | |
153 | AV_PIX_FMT_YUV444P, AV_PIX_FMT_YUV422P, | |
154 | AV_PIX_FMT_YUV420P, AV_PIX_FMT_YUV411P, | |
155 | AV_PIX_FMT_YUV410P, AV_PIX_FMT_YUV440P, | |
156 | AV_PIX_FMT_YUVJ444P, AV_PIX_FMT_YUVJ422P, | |
157 | AV_PIX_FMT_YUVJ420P, AV_PIX_FMT_YUVJ440P, | |
158 | AV_PIX_FMT_YUVA420P, | |
159 | AV_PIX_FMT_NONE | |
160 | }; | |
161 | ||
162 | ff_set_common_formats(ctx, ff_make_format_list(pix_fmts)); | |
163 | ||
164 | return 0; | |
165 | } | |
166 | ||
167 | static int config_input(AVFilterLink *inlink) | |
168 | { | |
169 | AVFilterContext *ctx = inlink->dst; | |
170 | DecimateContext *decimate = ctx->priv; | |
171 | const AVPixFmtDescriptor *pix_desc = av_pix_fmt_desc_get(inlink->format); | |
172 | decimate->hsub = pix_desc->log2_chroma_w; | |
173 | decimate->vsub = pix_desc->log2_chroma_h; | |
174 | ||
175 | return 0; | |
176 | } | |
177 | ||
178 | static int filter_frame(AVFilterLink *inlink, AVFrame *cur) | |
179 | { | |
180 | DecimateContext *decimate = inlink->dst->priv; | |
181 | AVFilterLink *outlink = inlink->dst->outputs[0]; | |
182 | int ret; | |
183 | ||
184 | if (decimate->ref && decimate_frame(inlink->dst, cur, decimate->ref)) { | |
185 | decimate->drop_count = FFMAX(1, decimate->drop_count+1); | |
186 | } else { | |
187 | av_frame_free(&decimate->ref); | |
188 | decimate->ref = cur; | |
189 | decimate->drop_count = FFMIN(-1, decimate->drop_count-1); | |
190 | ||
191 | if (ret = ff_filter_frame(outlink, av_frame_clone(cur)) < 0) | |
192 | return ret; | |
193 | } | |
194 | ||
195 | av_log(inlink->dst, AV_LOG_DEBUG, | |
196 | "%s pts:%s pts_time:%s drop_count:%d\n", | |
197 | decimate->drop_count > 0 ? "drop" : "keep", | |
198 | av_ts2str(cur->pts), av_ts2timestr(cur->pts, &inlink->time_base), | |
199 | decimate->drop_count); | |
200 | ||
201 | if (decimate->drop_count > 0) | |
202 | av_frame_free(&cur); | |
203 | ||
204 | return 0; | |
205 | } | |
206 | ||
207 | static int request_frame(AVFilterLink *outlink) | |
208 | { | |
209 | DecimateContext *decimate = outlink->src->priv; | |
210 | AVFilterLink *inlink = outlink->src->inputs[0]; | |
211 | int ret; | |
212 | ||
213 | do { | |
214 | ret = ff_request_frame(inlink); | |
215 | } while (decimate->drop_count > 0 && ret >= 0); | |
216 | ||
217 | return ret; | |
218 | } | |
219 | ||
220 | static const AVFilterPad mpdecimate_inputs[] = { | |
221 | { | |
222 | .name = "default", | |
223 | .type = AVMEDIA_TYPE_VIDEO, | |
224 | .config_props = config_input, | |
225 | .filter_frame = filter_frame, | |
226 | }, | |
227 | { NULL } | |
228 | }; | |
229 | ||
230 | static const AVFilterPad mpdecimate_outputs[] = { | |
231 | { | |
232 | .name = "default", | |
233 | .type = AVMEDIA_TYPE_VIDEO, | |
234 | .request_frame = request_frame, | |
235 | }, | |
236 | { NULL } | |
237 | }; | |
238 | ||
239 | AVFilter ff_vf_mpdecimate = { | |
240 | .name = "mpdecimate", | |
241 | .description = NULL_IF_CONFIG_SMALL("Remove near-duplicate frames."), | |
242 | .init = init, | |
243 | .uninit = uninit, | |
244 | .priv_size = sizeof(DecimateContext), | |
245 | .priv_class = &mpdecimate_class, | |
246 | .query_formats = query_formats, | |
247 | .inputs = mpdecimate_inputs, | |
248 | .outputs = mpdecimate_outputs, | |
249 | }; |