blob: dc345114cde69a1caa13055e452ef9dbd602100d
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 | av_log(ctx, AV_LOG_DEBUG, "%d>=hi ", d); |
89 | return 1; |
90 | } |
91 | if (d > decimate->lo) { |
92 | c++; |
93 | if (c > t) { |
94 | av_log(ctx, AV_LOG_DEBUG, "lo:%d>=%d ", c, t); |
95 | return 1; |
96 | } |
97 | } |
98 | } |
99 | } |
100 | |
101 | av_log(ctx, AV_LOG_DEBUG, "lo:%d<%d ", c, t); |
102 | return 0; |
103 | } |
104 | |
105 | /** |
106 | * Tell if the frame should be decimated, for example if it is no much |
107 | * different with respect to the reference frame ref. |
108 | */ |
109 | static int decimate_frame(AVFilterContext *ctx, |
110 | AVFrame *cur, AVFrame *ref) |
111 | { |
112 | DecimateContext *decimate = ctx->priv; |
113 | int plane; |
114 | |
115 | if (decimate->max_drop_count > 0 && |
116 | decimate->drop_count >= decimate->max_drop_count) |
117 | return 0; |
118 | if (decimate->max_drop_count < 0 && |
119 | (decimate->drop_count-1) > decimate->max_drop_count) |
120 | return 0; |
121 | |
122 | for (plane = 0; ref->data[plane] && ref->linesize[plane]; plane++) { |
123 | /* use 8x8 SAD even on subsampled planes. The blocks won't match up with |
124 | * luma blocks, but hopefully nobody is depending on this to catch |
125 | * localized chroma changes that wouldn't exceed the thresholds when |
126 | * diluted by using what's effectively a larger block size. |
127 | */ |
128 | int vsub = plane == 1 || plane == 2 ? decimate->vsub : 0; |
129 | int hsub = plane == 1 || plane == 2 ? decimate->hsub : 0; |
130 | if (diff_planes(ctx, |
131 | cur->data[plane], cur->linesize[plane], |
132 | ref->data[plane], ref->linesize[plane], |
133 | AV_CEIL_RSHIFT(ref->width, hsub), |
134 | AV_CEIL_RSHIFT(ref->height, vsub))) { |
135 | emms_c(); |
136 | return 0; |
137 | } |
138 | } |
139 | |
140 | emms_c(); |
141 | return 1; |
142 | } |
143 | |
144 | static av_cold int init(AVFilterContext *ctx) |
145 | { |
146 | DecimateContext *decimate = ctx->priv; |
147 | |
148 | decimate->sad = av_pixelutils_get_sad_fn(3, 3, 0, ctx); // 8x8, not aligned on blocksize |
149 | if (!decimate->sad) |
150 | return AVERROR(EINVAL); |
151 | |
152 | av_log(ctx, AV_LOG_VERBOSE, "max_drop_count:%d hi:%d lo:%d frac:%f\n", |
153 | decimate->max_drop_count, decimate->hi, decimate->lo, decimate->frac); |
154 | |
155 | return 0; |
156 | } |
157 | |
158 | static av_cold void uninit(AVFilterContext *ctx) |
159 | { |
160 | DecimateContext *decimate = ctx->priv; |
161 | av_frame_free(&decimate->ref); |
162 | } |
163 | |
164 | static int query_formats(AVFilterContext *ctx) |
165 | { |
166 | static const enum AVPixelFormat pix_fmts[] = { |
167 | AV_PIX_FMT_YUV444P, AV_PIX_FMT_YUV422P, |
168 | AV_PIX_FMT_YUV420P, AV_PIX_FMT_YUV411P, |
169 | AV_PIX_FMT_YUV410P, AV_PIX_FMT_YUV440P, |
170 | AV_PIX_FMT_YUVJ444P, AV_PIX_FMT_YUVJ422P, |
171 | AV_PIX_FMT_YUVJ420P, AV_PIX_FMT_YUVJ440P, |
172 | AV_PIX_FMT_YUVA420P, |
173 | |
174 | AV_PIX_FMT_GBRP, |
175 | |
176 | AV_PIX_FMT_YUVA444P, |
177 | AV_PIX_FMT_YUVA422P, |
178 | |
179 | AV_PIX_FMT_NONE |
180 | }; |
181 | AVFilterFormats *fmts_list = ff_make_format_list(pix_fmts); |
182 | if (!fmts_list) |
183 | return AVERROR(ENOMEM); |
184 | return ff_set_common_formats(ctx, fmts_list); |
185 | } |
186 | |
187 | static int config_input(AVFilterLink *inlink) |
188 | { |
189 | AVFilterContext *ctx = inlink->dst; |
190 | DecimateContext *decimate = ctx->priv; |
191 | const AVPixFmtDescriptor *pix_desc = av_pix_fmt_desc_get(inlink->format); |
192 | decimate->hsub = pix_desc->log2_chroma_w; |
193 | decimate->vsub = pix_desc->log2_chroma_h; |
194 | |
195 | return 0; |
196 | } |
197 | |
198 | static int filter_frame(AVFilterLink *inlink, AVFrame *cur) |
199 | { |
200 | DecimateContext *decimate = inlink->dst->priv; |
201 | AVFilterLink *outlink = inlink->dst->outputs[0]; |
202 | int ret; |
203 | |
204 | if (decimate->ref && decimate_frame(inlink->dst, cur, decimate->ref)) { |
205 | decimate->drop_count = FFMAX(1, decimate->drop_count+1); |
206 | } else { |
207 | av_frame_free(&decimate->ref); |
208 | decimate->ref = cur; |
209 | decimate->drop_count = FFMIN(-1, decimate->drop_count-1); |
210 | |
211 | if ((ret = ff_filter_frame(outlink, av_frame_clone(cur))) < 0) |
212 | return ret; |
213 | } |
214 | |
215 | av_log(inlink->dst, AV_LOG_DEBUG, |
216 | "%s pts:%s pts_time:%s drop_count:%d\n", |
217 | decimate->drop_count > 0 ? "drop" : "keep", |
218 | av_ts2str(cur->pts), av_ts2timestr(cur->pts, &inlink->time_base), |
219 | decimate->drop_count); |
220 | |
221 | if (decimate->drop_count > 0) |
222 | av_frame_free(&cur); |
223 | |
224 | return 0; |
225 | } |
226 | |
227 | static const AVFilterPad mpdecimate_inputs[] = { |
228 | { |
229 | .name = "default", |
230 | .type = AVMEDIA_TYPE_VIDEO, |
231 | .config_props = config_input, |
232 | .filter_frame = filter_frame, |
233 | }, |
234 | { NULL } |
235 | }; |
236 | |
237 | static const AVFilterPad mpdecimate_outputs[] = { |
238 | { |
239 | .name = "default", |
240 | .type = AVMEDIA_TYPE_VIDEO, |
241 | }, |
242 | { NULL } |
243 | }; |
244 | |
245 | AVFilter ff_vf_mpdecimate = { |
246 | .name = "mpdecimate", |
247 | .description = NULL_IF_CONFIG_SMALL("Remove near-duplicate frames."), |
248 | .init = init, |
249 | .uninit = uninit, |
250 | .priv_size = sizeof(DecimateContext), |
251 | .priv_class = &mpdecimate_class, |
252 | .query_formats = query_formats, |
253 | .inputs = mpdecimate_inputs, |
254 | .outputs = mpdecimate_outputs, |
255 | }; |
256 |