[FFmpeg-devel] libavfilter: vid.stab filters patch for 0.98 version
Stefano Sabatini
stefasab at gmail.com
Tue Jan 14 12:37:56 CET 2014
On date Saturday 2014-01-04 15:46:29 +0100, Georg Martius encoded:
> Hi all,
>
> please find attached the patch for vf_vidstab* in libavfilter to work with the
> newest version of the vid.stab library.
>
> Regards,
> Georg
> --
> ---- Georg Martius, Tel: +49 177 6413311 -----
> --------- http://georg.hronopik.de -------------
> From af046f1ea7731e264e2815e8ed026f1b33dea374 Mon Sep 17 00:00:00 2001
> From: Georg Martius <martius at mis.mpg.de>
> Date: Tue, 8 Oct 2013 23:57:17 +0200
> Subject: [PATCH] vidstab filters: changes for vid.stab 0.98
> default value for accuracy is 15 (max)
> zoomspeed for dynamic zoom (optzoom=2)
> camera path optimization algorithm configurable
> optzoom documenation in code updated
> use new localmotions calculation function
> commandline arguments: ranges and documentation improved
> debug mode option in vf_vidstabtransform
>
> Signed-off-by: Georg Martius <martius at mis.mpg.de>
> ---
> doc/filters.texi | 28 ++++++++++++++++++++++++++--
> libavfilter/vf_vidstabdetect.c | 3 ++-
> libavfilter/vf_vidstabtransform.c | 34 ++++++++++++++++++++++++++++------
> 3 files changed, 56 insertions(+), 9 deletions(-)
>
> diff --git a/doc/filters.texi b/doc/filters.texi
> index 0bf945b..93739ea 100644
> --- a/doc/filters.texi
> +++ b/doc/filters.texi
> @@ -7717,7 +7717,7 @@ value of 10 means strong shakiness. Default value is 5.
> @item accuracy
> Set the accuracy of the detection process. It must be a value in the
> range 1-15. A value of 1 means low accuracy, a value of 15 means high
> -accuracy. Default value is 9.
> +accuracy. Default value is 15.
>
> @item stepsize
> Set stepsize of the search process. The region around minimum is
> @@ -7802,6 +7802,18 @@ number of frames (value*2 + 1) used for lowpass filtering the camera movements
> (10 in the past and 10 in the future) to smoothen the motion in the
> video. A larger values leads to a smoother video, but limits the
> acceleration of the camera (pan/tilt movements).
> +0 is a special case where a static camera is simulated.
> +
> + at item optalgo
> +camera path optimization algorithm:
Set camera path ...
> + at table @samp
> + at c @item opt
> + at c optimal camera path w.r.t. L1 constraints (default)
why the comment?
> + at item gauss
> +gaussian kernel low-pass filter on camera motion (default)
> + at item avg
> +averaging on transformations
> + at end table
>
> @item maxshift
> maximal number of pixels to translate frames (default: -1 no limit)
> @@ -7855,11 +7867,14 @@ disabled
> @item 1
> optimal static zoom value is determined (only very strong movements will lead to visible borders) (default)
> @item 2
> -optimal adaptive zoom value is determined (no borders will be visible)
> +optimal adaptive zoom value is determined (no borders will be visible), see zoomspeed
@option{zoomspeed}
> @end table
> Note that the value given at zoom is added to the one calculated
> here.
>
> + at item zoomspeed
> +for adative zoom: percent to zoom maximally each frame (default 0.2%)
Set percent to zoom ...
Range is from 0 to 5, default value is 0.25.
> +
> @item interpol
> type of interpolation
>
> @@ -7886,6 +7901,15 @@ off (default)
> virtual tripod mode: equivalent to @code{relative=0:smoothing=0}
> @end table
>
> + at item debug
> +debug more: verbose and storing of global motions to file
Increase log verbosity if set to 1. ... Default value is 0.
"verbose and storing of global motions to file" -> this is not valid
English
> + at table @samp
> + at item 0
> +disabled (default)
> + at item 1
> +enabled
> + at end table
> +
> @end table
>
> @subsection Examples
> diff --git a/libavfilter/vf_vidstabdetect.c b/libavfilter/vf_vidstabdetect.c
> index 99d99ce..e7463cc 100644
> --- a/libavfilter/vf_vidstabdetect.c
> +++ b/libavfilter/vf_vidstabdetect.c
> @@ -49,7 +49,7 @@ static const AVOption vidstabdetect_options[] = {
> {"result", "path to the file used to write the transforms", OFFSET(result), AV_OPT_TYPE_STRING, {.str = DEFAULT_RESULT_NAME}, .flags = FLAGS},
> {"shakiness", "how shaky is the video and how quick is the camera?"
> " 1: little (fast) 10: very strong/quick (slow)", OFFSETC(shakiness), AV_OPT_TYPE_INT, {.i64 = 5}, 1, 10, FLAGS},
> - {"accuracy", "(>=shakiness) 1: low 15: high (slow)", OFFSETC(accuracy), AV_OPT_TYPE_INT, {.i64 = 9}, 1, 15, FLAGS},
> + {"accuracy", "(>=shakiness) 1: low 15: high (slow)", OFFSETC(accuracy), AV_OPT_TYPE_INT, {.i64 = 15}, 1, 15, FLAGS},
> {"stepsize", "region around minimum is scanned with 1 pixel resolution", OFFSETC(stepSize), AV_OPT_TYPE_INT, {.i64 = 6}, 1, 32, FLAGS},
> {"mincontrast", "below this contrast a field is discarded (0-1)", OFFSETC(contrastThreshold), AV_OPT_TYPE_DOUBLE, {.dbl = 0.25}, 0.0, 1.0, FLAGS},
> {"show", "0: draw nothing; 1,2: show fields and transforms", OFFSETC(show), AV_OPT_TYPE_INT, {.i64 = 0}, 0, 2, FLAGS},
> @@ -135,6 +135,7 @@ static int config_input(AVFilterLink *inlink)
> av_log(ctx, AV_LOG_INFO, " accuracy = %d\n", sd->conf.accuracy);
> av_log(ctx, AV_LOG_INFO, " stepsize = %d\n", sd->conf.stepSize);
> av_log(ctx, AV_LOG_INFO, " mincontrast = %f\n", sd->conf.contrastThreshold);
> + av_log(ctx, AV_LOG_INFO, " tripod = %d\n", sd->conf.virtualTripod);
> av_log(ctx, AV_LOG_INFO, " show = %d\n", sd->conf.show);
> av_log(ctx, AV_LOG_INFO, " result = %s\n", sd->result);
>
> diff --git a/libavfilter/vf_vidstabtransform.c b/libavfilter/vf_vidstabtransform.c
> index 40c0593..d8b1e40 100644
> --- a/libavfilter/vf_vidstabtransform.c
> +++ b/libavfilter/vf_vidstabtransform.c
> @@ -39,6 +39,7 @@ typedef struct {
> VSTransformations trans; // transformations
> char *input; // name of transform file
> int tripod;
> + int debug;
> } TransformContext;
>
> #define OFFSET(x) offsetof(TransformContext, x)
> @@ -49,7 +50,15 @@ static const AVOption vidstabtransform_options[] = {
> {"input", "path to the file storing the transforms", OFFSET(input),
> AV_OPT_TYPE_STRING, {.str = DEFAULT_INPUT_NAME}, .flags = FLAGS },
> {"smoothing", "number of frames*2 + 1 used for lowpass filtering", OFFSETC(smoothing),
> - AV_OPT_TYPE_INT, {.i64 = 10}, 1, 1000, FLAGS},
> + AV_OPT_TYPE_INT, {.i64 = 15}, 0, 1000, FLAGS},
> + {"optalgo", "camera path optimization algo", OFFSETC(camPathAlgo),
> + AV_OPT_TYPE_INT, {.i64 = VSOptimalL1}, VSOptimalL1, VSAvg, FLAGS, "optalgo"},
> + { "opt", "Global optimization", 0,
_g_lobal optimization
> + AV_OPT_TYPE_CONST, {.i64 = VSOptimalL1 }, 0, 0, FLAGS, "optalgo"},
> + { "gauss", "gaussian kernel", 0,
> + AV_OPT_TYPE_CONST, {.i64 = VSGaussian }, 0, 0, FLAGS, "optalgo"},
> + { "avg", "simple averaging on motion", 0,
> + AV_OPT_TYPE_CONST, {.i64 = VSAvg }, 0, 0, FLAGS, "optalgo"},
> {"maxshift", "maximal number of pixels to translate image", OFFSETC(maxShift),
> AV_OPT_TYPE_INT, {.i64 = -1}, -1, 500, FLAGS},
> {"maxangle", "maximal angle in rad to rotate image", OFFSETC(maxAngle),
> @@ -66,8 +75,10 @@ static const AVOption vidstabtransform_options[] = {
> AV_OPT_TYPE_INT, {.i64 = 1}, 0, 1, FLAGS},
> {"zoom", "percentage to zoom >0: zoom in, <0 zoom out", OFFSETC(zoom),
> AV_OPT_TYPE_DOUBLE, {.dbl = 0}, -100, 100, FLAGS},
> - {"optzoom", "0: nothing, 1: determine optimal zoom (added to 'zoom')", OFFSETC(optZoom),
> + {"optzoom", "0: nothing, 1: optimal static zoom, 2: optimal dynamic zoom", OFFSETC(optZoom),
> AV_OPT_TYPE_INT, {.i64 = 1}, 0, 2, FLAGS},
> + {"zoomspeed", "for adative zoom: percent to zoom maximally each frame", OFFSETC(zoomSpeed),
> + AV_OPT_TYPE_DOUBLE, {.dbl = 0.25}, 0, 5, FLAGS},
> {"interpol", "type of interpolation", OFFSETC(interpolType),
> AV_OPT_TYPE_INT, {.i64 = 2}, 0, 3, FLAGS, "interpol"},
> { "no", "no interpolation", 0,
> @@ -80,6 +91,8 @@ static const AVOption vidstabtransform_options[] = {
> AV_OPT_TYPE_CONST, {.i64 = VS_BiCubic },0, 0, FLAGS, "interpol"},
> {"tripod", "if 1: virtual tripod mode (equiv. to relative=0:smoothing=0)", OFFSET(tripod),
> AV_OPT_TYPE_INT, {.i64 = 0}, 0, 1, FLAGS},
> + {"debug", "if 1: more output printed and global motions are stored to file",OFFSET(debug),
> + AV_OPT_TYPE_INT, {.i64 = 0}, 0, 1, FLAGS},
> {NULL}
> };
>
> @@ -153,12 +166,15 @@ static int config_input(AVFilterLink *inlink)
>
> // set values that are not initializes by the options
> tc->conf.modName = "vidstabtransform";
> - tc->conf.verbose =1;
> + tc->conf.verbose = 1 + tc->debug;
> if (tc->tripod) {
> - av_log(ctx, AV_LOG_INFO, "Virtual tripod mode: relative=0, smoothing=0");
> + av_log(ctx, AV_LOG_INFO, "Virtual tripod mode: relative=0, smoothing=0\n");
> tc->conf.relative = 0;
> tc->conf.smoothing = 0;
> }
> + tc->conf.simpleMotionCalculation = 0;
> + tc->conf.storeTransforms = tc->debug;
> + tc->conf.smoothZoom = 0;
>
> if (vsTransformDataInit(td, &tc->conf, &fi_src, &fi_dest) != VS_OK) {
> av_log(ctx, AV_LOG_ERROR, "initialization of vid.stab transform failed, please report a BUG\n");
> @@ -169,13 +185,19 @@ static int config_input(AVFilterLink *inlink)
> av_log(ctx, AV_LOG_INFO, "Video transformation/stabilization settings (pass 2/2):\n");
> av_log(ctx, AV_LOG_INFO, " input = %s\n", tc->input);
> av_log(ctx, AV_LOG_INFO, " smoothing = %d\n", tc->conf.smoothing);
> + av_log(ctx, AV_LOG_INFO, " optalgo = %s\n",
> + tc->conf.camPathAlgo==VSOptimalL1 ? "opt" :
> + (tc->conf.camPathAlgo == VSGaussian ? "gauss" : "avg" ));
nit: xxx == yyy;
> av_log(ctx, AV_LOG_INFO, " maxshift = %d\n", tc->conf.maxShift);
> av_log(ctx, AV_LOG_INFO, " maxangle = %f\n", tc->conf.maxAngle);
> av_log(ctx, AV_LOG_INFO, " crop = %s\n", tc->conf.crop ? "Black" : "Keep");
> av_log(ctx, AV_LOG_INFO, " relative = %s\n", tc->conf.relative ? "True": "False");
> av_log(ctx, AV_LOG_INFO, " invert = %s\n", tc->conf.invert ? "True" : "False");
> av_log(ctx, AV_LOG_INFO, " zoom = %f\n", tc->conf.zoom);
> - av_log(ctx, AV_LOG_INFO, " optzoom = %s\n", tc->conf.optZoom ? "On" : "Off");
> + av_log(ctx, AV_LOG_INFO, " optzoom = %s\n",
> + tc->conf.optZoom==1 ? "Static (1)" : (tc->conf.optZoom ==2 ? "Dynamic (2)" : "Off (0)" ));
> + if(tc->conf.optZoom==2)
if_(... == 2)
> + av_log(ctx, AV_LOG_INFO, " zoomspeed = %g\n", tc->conf.zoomSpeed );
weird indent
> av_log(ctx, AV_LOG_INFO, " interpol = %s\n", getInterpolationTypeName(tc->conf.interpolType));
>
> f = fopen(tc->input, "r");
> @@ -186,7 +208,7 @@ static int config_input(AVFilterLink *inlink)
> VSManyLocalMotions mlms;
> if (vsReadLocalMotionsFile(f, &mlms) == VS_OK) {
> // calculate the actual transforms from the local motions
> - if (vsLocalmotions2TransformsSimple(td, &mlms, &tc->trans) != VS_OK) {
> + if (vsLocalmotions2Transforms(td, &mlms, &tc->trans) != VS_OK) {
> av_log(ctx, AV_LOG_ERROR, "calculating transformations failed\n");
> return AVERROR(EINVAL);
> }
Should be good otherwise.
--
FFmpeg = Faithful and Foolish Merciful Pitiless Ecumenical God
More information about the ffmpeg-devel
mailing list