2 * Copyright (c) 2013 Georg Martius <georg dot martius at web dot de>
4 * This file is part of FFmpeg.
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.
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.
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
21 #define DEFAULT_INPUT_NAME "transforms.trf"
23 #include <vid.stab/libvidstab.h>
25 #include "libavutil/common.h"
26 #include "libavutil/opt.h"
27 #include "libavutil/imgutils.h"
31 #include "vidstabutils.h"
37 VSTransformConfig conf;
39 VSTransformations trans; // transformations
40 char *input; // name of transform file
45 #define OFFSET(x) offsetof(TransformContext, x)
46 #define OFFSETC(x) (offsetof(TransformContext, conf)+offsetof(VSTransformConfig, x))
47 #define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM
49 static const AVOption vidstabtransform_options[] = {
50 {"input", "path to the file storing the transforms", OFFSET(input),
51 AV_OPT_TYPE_STRING, {.str = DEFAULT_INPUT_NAME}, .flags = FLAGS },
52 {"smoothing", "number of frames*2 + 1 used for lowpass filtering", OFFSETC(smoothing),
53 AV_OPT_TYPE_INT, {.i64 = 15}, 0, 1000, FLAGS},
54 {"optalgo", "camera path optimization algo", OFFSETC(camPathAlgo),
55 AV_OPT_TYPE_INT, {.i64 = VSOptimalL1}, VSOptimalL1, VSAvg, FLAGS, "optalgo"},
56 { "opt", "global optimization", 0, // from version 1.0 on
57 AV_OPT_TYPE_CONST, {.i64 = VSOptimalL1 }, 0, 0, FLAGS, "optalgo"},
58 { "gauss", "gaussian kernel", 0,
59 AV_OPT_TYPE_CONST, {.i64 = VSGaussian }, 0, 0, FLAGS, "optalgo"},
60 { "avg", "simple averaging on motion", 0,
61 AV_OPT_TYPE_CONST, {.i64 = VSAvg }, 0, 0, FLAGS, "optalgo"},
62 {"maxshift", "maximal number of pixels to translate image", OFFSETC(maxShift),
63 AV_OPT_TYPE_INT, {.i64 = -1}, -1, 500, FLAGS},
64 {"maxangle", "maximal angle in rad to rotate image", OFFSETC(maxAngle),
65 AV_OPT_TYPE_DOUBLE, {.dbl = -1.0}, -1.0, 3.14, FLAGS},
66 {"crop", "set cropping mode", OFFSETC(crop),
67 AV_OPT_TYPE_INT, {.i64 = 0}, 0, 1, FLAGS, "crop"},
68 { "keep", "keep border", 0,
69 AV_OPT_TYPE_CONST, {.i64 = VSKeepBorder }, 0, 0, FLAGS, "crop"},
70 { "black", "black border", 0,
71 AV_OPT_TYPE_CONST, {.i64 = VSCropBorder }, 0, 0, FLAGS, "crop"},
72 {"invert", "1: invert transforms", OFFSETC(invert),
73 AV_OPT_TYPE_INT, {.i64 = 0}, 0, 1, FLAGS},
74 {"relative", "consider transforms as 0: absolute, 1: relative", OFFSETC(relative),
75 AV_OPT_TYPE_INT, {.i64 = 1}, 0, 1, FLAGS},
76 {"zoom", "percentage to zoom >0: zoom in, <0 zoom out", OFFSETC(zoom),
77 AV_OPT_TYPE_DOUBLE, {.dbl = 0}, -100, 100, FLAGS},
78 {"optzoom", "0: nothing, 1: optimal static zoom, 2: optimal dynamic zoom", OFFSETC(optZoom),
79 AV_OPT_TYPE_INT, {.i64 = 1}, 0, 2, FLAGS},
80 {"zoomspeed", "for adative zoom: percent to zoom maximally each frame", OFFSETC(zoomSpeed),
81 AV_OPT_TYPE_DOUBLE, {.dbl = 0.25}, 0, 5, FLAGS},
82 {"interpol", "type of interpolation", OFFSETC(interpolType),
83 AV_OPT_TYPE_INT, {.i64 = 2}, 0, 3, FLAGS, "interpol"},
84 { "no", "no interpolation", 0,
85 AV_OPT_TYPE_CONST, {.i64 = VS_Zero }, 0, 0, FLAGS, "interpol"},
86 { "linear", "linear (horizontal)", 0,
87 AV_OPT_TYPE_CONST, {.i64 = VS_Linear }, 0, 0, FLAGS, "interpol"},
88 { "bilinear","bi-linear", 0,
89 AV_OPT_TYPE_CONST, {.i64 = VS_BiLinear},0, 0, FLAGS, "interpol"},
90 { "bicubic", "bi-cubic", 0,
91 AV_OPT_TYPE_CONST, {.i64 = VS_BiCubic },0, 0, FLAGS, "interpol"},
92 {"tripod", "if 1: virtual tripod mode (equiv. to relative=0:smoothing=0)", OFFSET(tripod),
93 AV_OPT_TYPE_INT, {.i64 = 0}, 0, 1, FLAGS},
94 {"debug", "if 1: more output printed and global motions are stored to file",OFFSET(debug),
95 AV_OPT_TYPE_INT, {.i64 = 0}, 0, 1, FLAGS},
99 AVFILTER_DEFINE_CLASS(vidstabtransform);
101 static av_cold int init(AVFilterContext *ctx)
103 TransformContext *tc = ctx->priv;
104 vs_set_mem_and_log_functions();
105 tc->class = &vidstabtransform_class;
106 av_log(ctx, AV_LOG_VERBOSE, "vidstabtransform filter: init %s\n", LIBVIDSTAB_VERSION);
110 static av_cold void uninit(AVFilterContext *ctx)
112 TransformContext *tc = ctx->priv;
114 vsTransformDataCleanup(&tc->td);
115 vsTransformationsCleanup(&tc->trans);
118 static int query_formats(AVFilterContext *ctx)
120 // If you add something here also add it in vidstabutils.c
121 static const enum AVPixelFormat pix_fmts[] = {
122 AV_PIX_FMT_YUV444P, AV_PIX_FMT_YUV422P, AV_PIX_FMT_YUV420P,
123 AV_PIX_FMT_YUV411P, AV_PIX_FMT_YUV410P, AV_PIX_FMT_YUVA420P,
124 AV_PIX_FMT_YUV440P, AV_PIX_FMT_GRAY8,
125 AV_PIX_FMT_RGB24, AV_PIX_FMT_BGR24, AV_PIX_FMT_RGBA,
129 ff_set_common_formats(ctx, ff_make_format_list(pix_fmts));
134 static int config_input(AVFilterLink *inlink)
136 AVFilterContext *ctx = inlink->dst;
137 TransformContext *tc = ctx->priv;
140 const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(inlink->format);
142 VSTransformData *td = &(tc->td);
147 if (!vsFrameInfoInit(&fi_src, inlink->w, inlink->h,
148 av_2_vs_pixel_format(ctx, inlink->format)) ||
149 !vsFrameInfoInit(&fi_dest, inlink->w, inlink->h,
150 av_2_vs_pixel_format(ctx, inlink->format))) {
151 av_log(ctx, AV_LOG_ERROR, "unknown pixel format: %i (%s)",
152 inlink->format, desc->name);
153 return AVERROR(EINVAL);
156 if (fi_src.bytesPerPixel != av_get_bits_per_pixel(desc)/8 ||
157 fi_src.log2ChromaW != desc->log2_chroma_w ||
158 fi_src.log2ChromaH != desc->log2_chroma_h) {
159 av_log(ctx, AV_LOG_ERROR, "pixel-format error: bpp %i<>%i ",
160 fi_src.bytesPerPixel, av_get_bits_per_pixel(desc)/8);
161 av_log(ctx, AV_LOG_ERROR, "chroma_subsampl: w: %i<>%i h: %i<>%i\n",
162 fi_src.log2ChromaW, desc->log2_chroma_w,
163 fi_src.log2ChromaH, desc->log2_chroma_h);
164 return AVERROR(EINVAL);
167 // set values that are not initializes by the options
168 tc->conf.modName = "vidstabtransform";
169 tc->conf.verbose = 1 + tc->debug;
171 av_log(ctx, AV_LOG_INFO, "Virtual tripod mode: relative=0, smoothing=0\n");
172 tc->conf.relative = 0;
173 tc->conf.smoothing = 0;
175 tc->conf.simpleMotionCalculation = 0;
176 tc->conf.storeTransforms = tc->debug;
177 tc->conf.smoothZoom = 0;
179 if (vsTransformDataInit(td, &tc->conf, &fi_src, &fi_dest) != VS_OK) {
180 av_log(ctx, AV_LOG_ERROR, "initialization of vid.stab transform failed, please report a BUG\n");
181 return AVERROR(EINVAL);
184 vsTransformGetConfig(&tc->conf, td);
185 av_log(ctx, AV_LOG_INFO, "Video transformation/stabilization settings (pass 2/2):\n");
186 av_log(ctx, AV_LOG_INFO, " input = %s\n", tc->input);
187 av_log(ctx, AV_LOG_INFO, " smoothing = %d\n", tc->conf.smoothing);
188 av_log(ctx, AV_LOG_INFO, " optalgo = %s\n",
189 tc->conf.camPathAlgo == VSOptimalL1 ? "opt" :
190 (tc->conf.camPathAlgo == VSGaussian ? "gauss" : "avg"));
191 av_log(ctx, AV_LOG_INFO, " maxshift = %d\n", tc->conf.maxShift);
192 av_log(ctx, AV_LOG_INFO, " maxangle = %f\n", tc->conf.maxAngle);
193 av_log(ctx, AV_LOG_INFO, " crop = %s\n", tc->conf.crop ? "Black" : "Keep");
194 av_log(ctx, AV_LOG_INFO, " relative = %s\n", tc->conf.relative ? "True": "False");
195 av_log(ctx, AV_LOG_INFO, " invert = %s\n", tc->conf.invert ? "True" : "False");
196 av_log(ctx, AV_LOG_INFO, " zoom = %f\n", tc->conf.zoom);
197 av_log(ctx, AV_LOG_INFO, " optzoom = %s\n",
198 tc->conf.optZoom == 1 ? "Static (1)" : (tc->conf.optZoom == 2 ? "Dynamic (2)" : "Off (0)"));
199 if (tc->conf.optZoom == 2)
200 av_log(ctx, AV_LOG_INFO, " zoomspeed = %g\n", tc->conf.zoomSpeed);
201 av_log(ctx, AV_LOG_INFO, " interpol = %s\n", getInterpolationTypeName(tc->conf.interpolType));
203 f = fopen(tc->input, "r");
205 av_log(ctx, AV_LOG_ERROR, "cannot open input file %s\n", tc->input);
206 return AVERROR(errno);
208 VSManyLocalMotions mlms;
209 if (vsReadLocalMotionsFile(f, &mlms) == VS_OK) {
210 // calculate the actual transforms from the local motions
211 if (vsLocalmotions2Transforms(td, &mlms, &tc->trans) != VS_OK) {
212 av_log(ctx, AV_LOG_ERROR, "calculating transformations failed\n");
213 return AVERROR(EINVAL);
215 } else { // try to read old format
216 if (!vsReadOldTransforms(td, f, &tc->trans)) { /* read input file */
217 av_log(ctx, AV_LOG_ERROR, "error parsing input file %s\n", tc->input);
218 return AVERROR(EINVAL);
224 if (vsPreprocessTransforms(td, &tc->trans) != VS_OK) {
225 av_log(ctx, AV_LOG_ERROR, "error while preprocessing transforms\n");
226 return AVERROR(EINVAL);
229 // TODO: add sharpening, so far the user needs to call the unsharp filter manually
234 static int filter_frame(AVFilterLink *inlink, AVFrame *in)
236 AVFilterContext *ctx = inlink->dst;
237 TransformContext *tc = ctx->priv;
238 VSTransformData* td = &(tc->td);
240 AVFilterLink *outlink = inlink->dst->outputs[0];
246 if (av_frame_is_writable(in)) {
250 out = ff_get_video_buffer(outlink, outlink->w, outlink->h);
253 return AVERROR(ENOMEM);
255 av_frame_copy_props(out, in);
258 for (plane = 0; plane < vsTransformGetSrcFrameInfo(td)->planes; plane++) {
259 inframe.data[plane] = in->data[plane];
260 inframe.linesize[plane] = in->linesize[plane];
263 vsTransformPrepare(td, &inframe, &inframe);
264 } else { // separate frames
266 for (plane = 0; plane < vsTransformGetDestFrameInfo(td)->planes; plane++) {
267 outframe.data[plane] = out->data[plane];
268 outframe.linesize[plane] = out->linesize[plane];
270 vsTransformPrepare(td, &inframe, &outframe);
273 vsDoTransform(td, vsGetNextTransform(td, &tc->trans));
275 vsTransformFinish(td);
280 return ff_filter_frame(outlink, out);
283 static const AVFilterPad avfilter_vf_vidstabtransform_inputs[] = {
286 .type = AVMEDIA_TYPE_VIDEO,
287 .filter_frame = filter_frame,
288 .config_props = config_input,
293 static const AVFilterPad avfilter_vf_vidstabtransform_outputs[] = {
296 .type = AVMEDIA_TYPE_VIDEO,
301 AVFilter ff_vf_vidstabtransform = {
302 .name = "vidstabtransform",
303 .description = NULL_IF_CONFIG_SMALL("Transform the frames, "
304 "pass 2 of 2 for stabilization "
305 "(see vidstabdetect for pass 1)."),
306 .priv_size = sizeof(TransformContext),
309 .query_formats = query_formats,
310 .inputs = avfilter_vf_vidstabtransform_inputs,
311 .outputs = avfilter_vf_vidstabtransform_outputs,
312 .priv_class = &vidstabtransform_class,