lavfi/select: add scene detection.
Based on the shotdetect algorithm (http://shotdetect.nonutc.fr/ by Johan MATHE johan.mathe tremplin-utc.net).
This commit is contained in:
parent
6b87f0707d
commit
72868144e5
@ -3,6 +3,7 @@ releases are sorted from youngest to oldest.
|
|||||||
|
|
||||||
version next:
|
version next:
|
||||||
- INI output in ffprobe
|
- INI output in ffprobe
|
||||||
|
- Scene detection in libavfilter
|
||||||
|
|
||||||
|
|
||||||
version 0.11:
|
version 0.11:
|
||||||
|
1
configure
vendored
1
configure
vendored
@ -1708,6 +1708,7 @@ ocv_filter_deps="libopencv"
|
|||||||
pan_filter_deps="swresample"
|
pan_filter_deps="swresample"
|
||||||
removelogo_filter_deps="avcodec avformat swscale"
|
removelogo_filter_deps="avcodec avformat swscale"
|
||||||
scale_filter_deps="swscale"
|
scale_filter_deps="swscale"
|
||||||
|
select_filter_deps="avcodec"
|
||||||
super2xsai_filter_deps="gpl"
|
super2xsai_filter_deps="gpl"
|
||||||
tinterlace_filter_deps="gpl"
|
tinterlace_filter_deps="gpl"
|
||||||
yadif_filter_deps="gpl"
|
yadif_filter_deps="gpl"
|
||||||
|
@ -2603,6 +2603,12 @@ the frame is bottom-field-first
|
|||||||
@item pos
|
@item pos
|
||||||
the position in the file of the filtered frame, -1 if the information
|
the position in the file of the filtered frame, -1 if the information
|
||||||
is not available (e.g. for synthetic video)
|
is not available (e.g. for synthetic video)
|
||||||
|
|
||||||
|
@item scene
|
||||||
|
value between 0 and 1 to indicate a new scene; a low value reflects a low
|
||||||
|
probability for the current frame to introduce a new scene, while a higher
|
||||||
|
value means the current frame is more likely to be one (see the example below)
|
||||||
|
|
||||||
@end table
|
@end table
|
||||||
|
|
||||||
The default value of the select expression is "1".
|
The default value of the select expression is "1".
|
||||||
@ -2635,6 +2641,15 @@ select='gte(t\,10)*lte(t\,20)*eq(pict_type\,I)'
|
|||||||
select='isnan(prev_selected_t)+gte(t-prev_selected_t\,10)'
|
select='isnan(prev_selected_t)+gte(t-prev_selected_t\,10)'
|
||||||
@end example
|
@end example
|
||||||
|
|
||||||
|
Complete example to create a mosaic of the first scenes:
|
||||||
|
|
||||||
|
@example
|
||||||
|
ffmpeg -i video.avi -vf select='gt(scene\,0.4)',scale=160:120,tile -frames:v 1 preview.png
|
||||||
|
@end example
|
||||||
|
|
||||||
|
Comparing @var{scene} against a value between 0.3 and 0.5 is generally a sane
|
||||||
|
choice.
|
||||||
|
|
||||||
@section setdar, setsar
|
@section setdar, setsar
|
||||||
|
|
||||||
The @code{setdar} filter sets the Display Aspect Ratio for the filter
|
The @code{setdar} filter sets the Display Aspect Ratio for the filter
|
||||||
|
@ -25,7 +25,9 @@
|
|||||||
|
|
||||||
#include "libavutil/eval.h"
|
#include "libavutil/eval.h"
|
||||||
#include "libavutil/fifo.h"
|
#include "libavutil/fifo.h"
|
||||||
|
#include "libavcodec/dsputil.h"
|
||||||
#include "avfilter.h"
|
#include "avfilter.h"
|
||||||
|
#include "formats.h"
|
||||||
#include "video.h"
|
#include "video.h"
|
||||||
|
|
||||||
static const char *const var_names[] = {
|
static const char *const var_names[] = {
|
||||||
@ -62,6 +64,8 @@ static const char *const var_names[] = {
|
|||||||
"key", ///< tell if the frame is a key frame
|
"key", ///< tell if the frame is a key frame
|
||||||
"pos", ///< original position in the file of the frame
|
"pos", ///< original position in the file of the frame
|
||||||
|
|
||||||
|
"scene",
|
||||||
|
|
||||||
NULL
|
NULL
|
||||||
};
|
};
|
||||||
|
|
||||||
@ -99,6 +103,8 @@ enum var_name {
|
|||||||
VAR_KEY,
|
VAR_KEY,
|
||||||
VAR_POS,
|
VAR_POS,
|
||||||
|
|
||||||
|
VAR_SCENE,
|
||||||
|
|
||||||
VAR_VARS_NB
|
VAR_VARS_NB
|
||||||
};
|
};
|
||||||
|
|
||||||
@ -107,6 +113,11 @@ enum var_name {
|
|||||||
typedef struct {
|
typedef struct {
|
||||||
AVExpr *expr;
|
AVExpr *expr;
|
||||||
double var_values[VAR_VARS_NB];
|
double var_values[VAR_VARS_NB];
|
||||||
|
int do_scene_detect; ///< 1 if the expression requires scene detection variables, 0 otherwise
|
||||||
|
AVCodecContext *avctx; ///< codec context required for the DSPContext (scene detect only)
|
||||||
|
DSPContext c; ///< context providing optimized SAD methods (scene detect only)
|
||||||
|
double prev_mafd; ///< previous MAFD (scene detect only)
|
||||||
|
AVFilterBufferRef *prev_picref; ///< previous frame (scene detect only)
|
||||||
double select;
|
double select;
|
||||||
int cache_frames;
|
int cache_frames;
|
||||||
AVFifoBuffer *pending_frames; ///< FIFO buffer of video frames
|
AVFifoBuffer *pending_frames; ///< FIFO buffer of video frames
|
||||||
@ -128,6 +139,8 @@ static av_cold int init(AVFilterContext *ctx, const char *args, void *opaque)
|
|||||||
av_log(ctx, AV_LOG_ERROR, "Failed to allocate pending frames buffer.\n");
|
av_log(ctx, AV_LOG_ERROR, "Failed to allocate pending frames buffer.\n");
|
||||||
return AVERROR(ENOMEM);
|
return AVERROR(ENOMEM);
|
||||||
}
|
}
|
||||||
|
|
||||||
|
select->do_scene_detect = args && strstr(args, "scene");
|
||||||
return 0;
|
return 0;
|
||||||
}
|
}
|
||||||
|
|
||||||
@ -160,9 +173,49 @@ static int config_input(AVFilterLink *inlink)
|
|||||||
select->var_values[VAR_INTERLACE_TYPE_T] = INTERLACE_TYPE_T;
|
select->var_values[VAR_INTERLACE_TYPE_T] = INTERLACE_TYPE_T;
|
||||||
select->var_values[VAR_INTERLACE_TYPE_B] = INTERLACE_TYPE_B;
|
select->var_values[VAR_INTERLACE_TYPE_B] = INTERLACE_TYPE_B;
|
||||||
|
|
||||||
|
if (select->do_scene_detect) {
|
||||||
|
select->avctx = avcodec_alloc_context3(NULL);
|
||||||
|
if (!select->avctx)
|
||||||
|
return AVERROR(ENOMEM);
|
||||||
|
ff_dsputil_init(&select->c, select->avctx);
|
||||||
|
}
|
||||||
return 0;
|
return 0;
|
||||||
}
|
}
|
||||||
|
|
||||||
|
static double get_scene_score(AVFilterContext *ctx, AVFilterBufferRef *picref)
|
||||||
|
{
|
||||||
|
double ret = 0;
|
||||||
|
SelectContext *select = ctx->priv;
|
||||||
|
AVFilterBufferRef *prev_picref = select->prev_picref;
|
||||||
|
|
||||||
|
if (prev_picref &&
|
||||||
|
picref->video->h == prev_picref->video->h &&
|
||||||
|
picref->video->w == prev_picref->video->w &&
|
||||||
|
picref->linesize[0] == prev_picref->linesize[0]) {
|
||||||
|
int x, y;
|
||||||
|
int64_t sad;
|
||||||
|
double mafd, diff;
|
||||||
|
uint8_t *p1 = picref->data[0];
|
||||||
|
uint8_t *p2 = prev_picref->data[0];
|
||||||
|
const int linesize = picref->linesize[0];
|
||||||
|
|
||||||
|
for (sad = y = 0; y < picref->video->h; y += 8)
|
||||||
|
for (x = 0; x < linesize; x += 8)
|
||||||
|
sad += select->c.sad[1](select,
|
||||||
|
p1 + y * linesize + x,
|
||||||
|
p2 + y * linesize + x,
|
||||||
|
linesize, 8);
|
||||||
|
emms_c();
|
||||||
|
mafd = sad / (picref->video->h * picref->video->w * 3);
|
||||||
|
diff = llabs(mafd - select->prev_mafd);
|
||||||
|
ret = av_clipf(FFMIN(mafd, diff) / 100., 0, 1);
|
||||||
|
select->prev_mafd = mafd;
|
||||||
|
avfilter_unref_buffer(prev_picref);
|
||||||
|
}
|
||||||
|
select->prev_picref = avfilter_ref_buffer(picref, ~0);
|
||||||
|
return ret;
|
||||||
|
}
|
||||||
|
|
||||||
#define D2TS(d) (isnan(d) ? AV_NOPTS_VALUE : (int64_t)(d))
|
#define D2TS(d) (isnan(d) ? AV_NOPTS_VALUE : (int64_t)(d))
|
||||||
#define TS2D(ts) ((ts) == AV_NOPTS_VALUE ? NAN : (double)(ts))
|
#define TS2D(ts) ((ts) == AV_NOPTS_VALUE ? NAN : (double)(ts))
|
||||||
|
|
||||||
@ -172,6 +225,8 @@ static int select_frame(AVFilterContext *ctx, AVFilterBufferRef *picref)
|
|||||||
AVFilterLink *inlink = ctx->inputs[0];
|
AVFilterLink *inlink = ctx->inputs[0];
|
||||||
double res;
|
double res;
|
||||||
|
|
||||||
|
if (select->do_scene_detect)
|
||||||
|
select->var_values[VAR_SCENE] = get_scene_score(ctx, picref);
|
||||||
if (isnan(select->var_values[VAR_START_PTS]))
|
if (isnan(select->var_values[VAR_START_PTS]))
|
||||||
select->var_values[VAR_START_PTS] = TS2D(picref->pts);
|
select->var_values[VAR_START_PTS] = TS2D(picref->pts);
|
||||||
if (isnan(select->var_values[VAR_START_T]))
|
if (isnan(select->var_values[VAR_START_T]))
|
||||||
@ -315,6 +370,28 @@ static av_cold void uninit(AVFilterContext *ctx)
|
|||||||
avfilter_unref_buffer(picref);
|
avfilter_unref_buffer(picref);
|
||||||
av_fifo_free(select->pending_frames);
|
av_fifo_free(select->pending_frames);
|
||||||
select->pending_frames = NULL;
|
select->pending_frames = NULL;
|
||||||
|
|
||||||
|
if (select->do_scene_detect) {
|
||||||
|
avfilter_unref_bufferp(&select->prev_picref);
|
||||||
|
avcodec_close(select->avctx);
|
||||||
|
av_freep(&select->avctx);
|
||||||
|
}
|
||||||
|
}
|
||||||
|
|
||||||
|
static int query_formats(AVFilterContext *ctx)
|
||||||
|
{
|
||||||
|
SelectContext *select = ctx->priv;
|
||||||
|
|
||||||
|
if (!select->do_scene_detect) {
|
||||||
|
return ff_default_query_formats(ctx);
|
||||||
|
} else {
|
||||||
|
static const enum PixelFormat pix_fmts[] = {
|
||||||
|
PIX_FMT_RGB24, PIX_FMT_BGR24,
|
||||||
|
PIX_FMT_NONE
|
||||||
|
};
|
||||||
|
avfilter_set_common_pixel_formats(ctx, avfilter_make_format_list(pix_fmts));
|
||||||
|
}
|
||||||
|
return 0;
|
||||||
}
|
}
|
||||||
|
|
||||||
AVFilter avfilter_vf_select = {
|
AVFilter avfilter_vf_select = {
|
||||||
@ -322,6 +399,7 @@ AVFilter avfilter_vf_select = {
|
|||||||
.description = NULL_IF_CONFIG_SMALL("Select frames to pass in output."),
|
.description = NULL_IF_CONFIG_SMALL("Select frames to pass in output."),
|
||||||
.init = init,
|
.init = init,
|
||||||
.uninit = uninit,
|
.uninit = uninit,
|
||||||
|
.query_formats = query_formats,
|
||||||
|
|
||||||
.priv_size = sizeof(SelectContext),
|
.priv_size = sizeof(SelectContext),
|
||||||
|
|
||||||
|
Loading…
x
Reference in New Issue
Block a user