1
0
mirror of https://github.com/FFmpeg/FFmpeg.git synced 2024-12-23 12:43:46 +02:00

avfilter: add thistogram video filter

This commit is contained in:
Paul B Mahol 2019-12-26 11:38:53 +01:00
parent 755ad01dd1
commit cc43c2f29a
6 changed files with 166 additions and 25 deletions

View File

@ -28,6 +28,7 @@ version <next>:
- mvdv decoder - mvdv decoder
- mvha decoder - mvha decoder
- MPEG-H 3D Audio support in mp4 - MPEG-H 3D Audio support in mp4
- thistogram filter
version 4.2: version 4.2:

View File

@ -11694,6 +11694,7 @@ the histogram. Possible values are @code{none}, @code{weak} or
@code{strong}. It defaults to @code{none}. @code{strong}. It defaults to @code{none}.
@end table @end table
@anchor{histogram}
@section histogram @section histogram
Compute and draw a color distribution histogram for the input video. Compute and draw a color distribution histogram for the input video.
@ -17745,6 +17746,55 @@ PAL output (25i):
16p: 33333334 16p: 33333334
@end example @end example
@section thistogram
Compute and draw a color distribution histogram for the input video across time.
Unlike @ref{histogram} video filter which only shows histogram of single input frame
at certain time, this filter shows also past histograms of number of frames defined
by @code{width} option.
The computed histogram is a representation of the color component
distribution in an image.
The filter accepts the following options:
@table @option
@item width, w
Set width of single color component output. Default value is @code{0}.
Value of @code{0} means width will be picked from input video.
This also set number of passed histograms to keep.
Allowed range is [0, 8192].
@item display_mode, d
Set display mode.
It accepts the following values:
@table @samp
@item stack
Per color component graphs are placed below each other.
@item parade
Per color component graphs are placed side by side.
@item overlay
Presents information identical to that in the @code{parade}, except
that the graphs representing color components are superimposed directly
over one another.
@end table
Default is @code{stack}.
@item levels_mode, m
Set mode. Can be either @code{linear}, or @code{logarithmic}.
Default is @code{linear}.
@item components, c
Set what color components to display.
Default is @code{7}.
@item bgopacity, b
Set background opacity. Default is @code{0.5}.
@end table
@section threshold @section threshold
Apply threshold effect to video stream. Apply threshold effect to video stream.

View File

@ -401,6 +401,7 @@ OBJS-$(CONFIG_SWAPRECT_FILTER) += vf_swaprect.o
OBJS-$(CONFIG_SWAPUV_FILTER) += vf_swapuv.o OBJS-$(CONFIG_SWAPUV_FILTER) += vf_swapuv.o
OBJS-$(CONFIG_TBLEND_FILTER) += vf_blend.o framesync.o OBJS-$(CONFIG_TBLEND_FILTER) += vf_blend.o framesync.o
OBJS-$(CONFIG_TELECINE_FILTER) += vf_telecine.o OBJS-$(CONFIG_TELECINE_FILTER) += vf_telecine.o
OBJS-$(CONFIG_THISTOGRAM_FILTER) += vf_histogram.o
OBJS-$(CONFIG_THRESHOLD_FILTER) += vf_threshold.o framesync.o OBJS-$(CONFIG_THRESHOLD_FILTER) += vf_threshold.o framesync.o
OBJS-$(CONFIG_THUMBNAIL_FILTER) += vf_thumbnail.o OBJS-$(CONFIG_THUMBNAIL_FILTER) += vf_thumbnail.o
OBJS-$(CONFIG_THUMBNAIL_CUDA_FILTER) += vf_thumbnail_cuda.o vf_thumbnail_cuda.ptx.o OBJS-$(CONFIG_THUMBNAIL_CUDA_FILTER) += vf_thumbnail_cuda.o vf_thumbnail_cuda.ptx.o

View File

@ -382,6 +382,7 @@ extern AVFilter ff_vf_swaprect;
extern AVFilter ff_vf_swapuv; extern AVFilter ff_vf_swapuv;
extern AVFilter ff_vf_tblend; extern AVFilter ff_vf_tblend;
extern AVFilter ff_vf_telecine; extern AVFilter ff_vf_telecine;
extern AVFilter ff_vf_thistogram;
extern AVFilter ff_vf_threshold; extern AVFilter ff_vf_threshold;
extern AVFilter ff_vf_thumbnail; extern AVFilter ff_vf_thumbnail;
extern AVFilter ff_vf_thumbnail_cuda; extern AVFilter ff_vf_thumbnail_cuda;

View File

@ -30,8 +30,8 @@
#include "libavutil/version.h" #include "libavutil/version.h"
#define LIBAVFILTER_VERSION_MAJOR 7 #define LIBAVFILTER_VERSION_MAJOR 7
#define LIBAVFILTER_VERSION_MINOR 69 #define LIBAVFILTER_VERSION_MINOR 70
#define LIBAVFILTER_VERSION_MICRO 101 #define LIBAVFILTER_VERSION_MICRO 100
#define LIBAVFILTER_VERSION_INT AV_VERSION_INT(LIBAVFILTER_VERSION_MAJOR, \ #define LIBAVFILTER_VERSION_INT AV_VERSION_INT(LIBAVFILTER_VERSION_MAJOR, \

View File

@ -1,5 +1,5 @@
/* /*
* Copyright (c) 2012-2013 Paul B Mahol * Copyright (c) 2012-2019 Paul B Mahol
* *
* This file is part of FFmpeg. * This file is part of FFmpeg.
* *
@ -31,8 +31,11 @@
typedef struct HistogramContext { typedef struct HistogramContext {
const AVClass *class; ///< AVClass context for log and options purpose const AVClass *class; ///< AVClass context for log and options purpose
int thistogram;
unsigned histogram[256*256]; unsigned histogram[256*256];
int histogram_size; int histogram_size;
int width;
int x_pos;
int mult; int mult;
int ncomp; int ncomp;
int dncomp; int dncomp;
@ -48,25 +51,30 @@ typedef struct HistogramContext {
float bgopacity; float bgopacity;
int planewidth[4]; int planewidth[4];
int planeheight[4]; int planeheight[4];
int start[4];
AVFrame *out;
} HistogramContext; } HistogramContext;
#define OFFSET(x) offsetof(HistogramContext, x) #define OFFSET(x) offsetof(HistogramContext, x)
#define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM #define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM
#define COMMON_OPTIONS \
{ "display_mode", "set display mode", OFFSET(display_mode), AV_OPT_TYPE_INT, {.i64=2}, 0, 2, FLAGS, "display_mode"}, \
{ "d", "set display mode", OFFSET(display_mode), AV_OPT_TYPE_INT, {.i64=2}, 0, 2, FLAGS, "display_mode"}, \
{ "overlay", NULL, 0, AV_OPT_TYPE_CONST, {.i64=0}, 0, 0, FLAGS, "display_mode" }, \
{ "parade", NULL, 0, AV_OPT_TYPE_CONST, {.i64=1}, 0, 0, FLAGS, "display_mode" }, \
{ "stack", NULL, 0, AV_OPT_TYPE_CONST, {.i64=2}, 0, 0, FLAGS, "display_mode" }, \
{ "levels_mode", "set levels mode", OFFSET(levels_mode), AV_OPT_TYPE_INT, {.i64=0}, 0, 1, FLAGS, "levels_mode"}, \
{ "m", "set levels mode", OFFSET(levels_mode), AV_OPT_TYPE_INT, {.i64=0}, 0, 1, FLAGS, "levels_mode"}, \
{ "linear", NULL, 0, AV_OPT_TYPE_CONST, {.i64=0}, 0, 0, FLAGS, "levels_mode" }, \
{ "logarithmic", NULL, 0, AV_OPT_TYPE_CONST, {.i64=1}, 0, 0, FLAGS, "levels_mode" }, \
{ "components", "set color components to display", OFFSET(components), AV_OPT_TYPE_INT, {.i64=7}, 1, 15, FLAGS}, \
{ "c", "set color components to display", OFFSET(components), AV_OPT_TYPE_INT, {.i64=7}, 1, 15, FLAGS},
static const AVOption histogram_options[] = { static const AVOption histogram_options[] = {
{ "level_height", "set level height", OFFSET(level_height), AV_OPT_TYPE_INT, {.i64=200}, 50, 2048, FLAGS}, { "level_height", "set level height", OFFSET(level_height), AV_OPT_TYPE_INT, {.i64=200}, 50, 2048, FLAGS},
{ "scale_height", "set scale height", OFFSET(scale_height), AV_OPT_TYPE_INT, {.i64=12}, 0, 40, FLAGS}, { "scale_height", "set scale height", OFFSET(scale_height), AV_OPT_TYPE_INT, {.i64=12}, 0, 40, FLAGS},
{ "display_mode", "set display mode", OFFSET(display_mode), AV_OPT_TYPE_INT, {.i64=2}, 0, 2, FLAGS, "display_mode"}, COMMON_OPTIONS
{ "d", "set display mode", OFFSET(display_mode), AV_OPT_TYPE_INT, {.i64=2}, 0, 2, FLAGS, "display_mode"},
{ "overlay", NULL, 0, AV_OPT_TYPE_CONST, {.i64=0}, 0, 0, FLAGS, "display_mode" },
{ "parade", NULL, 0, AV_OPT_TYPE_CONST, {.i64=1}, 0, 0, FLAGS, "display_mode" },
{ "stack", NULL, 0, AV_OPT_TYPE_CONST, {.i64=2}, 0, 0, FLAGS, "display_mode" },
{ "levels_mode", "set levels mode", OFFSET(levels_mode), AV_OPT_TYPE_INT, {.i64=0}, 0, 1, FLAGS, "levels_mode"},
{ "m", "set levels mode", OFFSET(levels_mode), AV_OPT_TYPE_INT, {.i64=0}, 0, 1, FLAGS, "levels_mode"},
{ "linear", NULL, 0, AV_OPT_TYPE_CONST, {.i64=0}, 0, 0, FLAGS, "levels_mode" },
{ "logarithmic", NULL, 0, AV_OPT_TYPE_CONST, {.i64=1}, 0, 0, FLAGS, "levels_mode" },
{ "components", "set color components to display", OFFSET(components), AV_OPT_TYPE_INT, {.i64=7}, 1, 15, FLAGS},
{ "c", "set color components to display", OFFSET(components), AV_OPT_TYPE_INT, {.i64=7}, 1, 15, FLAGS},
{ "fgopacity", "set foreground opacity", OFFSET(fgopacity), AV_OPT_TYPE_FLOAT, {.dbl=0.7}, 0, 1, FLAGS}, { "fgopacity", "set foreground opacity", OFFSET(fgopacity), AV_OPT_TYPE_FLOAT, {.dbl=0.7}, 0, 1, FLAGS},
{ "f", "set foreground opacity", OFFSET(fgopacity), AV_OPT_TYPE_FLOAT, {.dbl=0.7}, 0, 1, FLAGS}, { "f", "set foreground opacity", OFFSET(fgopacity), AV_OPT_TYPE_FLOAT, {.dbl=0.7}, 0, 1, FLAGS},
{ "bgopacity", "set background opacity", OFFSET(bgopacity), AV_OPT_TYPE_FLOAT, {.dbl=0.5}, 0, 1, FLAGS}, { "bgopacity", "set background opacity", OFFSET(bgopacity), AV_OPT_TYPE_FLOAT, {.dbl=0.5}, 0, 1, FLAGS},
@ -209,10 +217,13 @@ static int config_input(AVFilterLink *inlink)
case AV_PIX_FMT_GBRP: case AV_PIX_FMT_GBRP:
memcpy(s->bg_color, black_gbrp_color, 4); memcpy(s->bg_color, black_gbrp_color, 4);
memcpy(s->fg_color, white_gbrp_color, 4); memcpy(s->fg_color, white_gbrp_color, 4);
s->start[0] = s->start[1] = s->start[2] = s->start[3] = 0;
break; break;
default: default:
memcpy(s->bg_color, black_yuva_color, 4); memcpy(s->bg_color, black_yuva_color, 4);
memcpy(s->fg_color, white_yuva_color, 4); memcpy(s->fg_color, white_yuva_color, 4);
s->start[0] = s->start[3] = 0;
s->start[1] = s->start[2] = s->histogram_size / 2;
} }
s->fg_color[3] = s->fgopacity * 255; s->fg_color[3] = s->fgopacity * 255;
@ -232,12 +243,23 @@ static int config_output(AVFilterLink *outlink)
HistogramContext *s = ctx->priv; HistogramContext *s = ctx->priv;
int ncomp = 0, i; int ncomp = 0, i;
if (!strcmp(ctx->filter->name, "thistogram"))
s->thistogram = 1;
for (i = 0; i < s->ncomp; i++) { for (i = 0; i < s->ncomp; i++) {
if ((1 << i) & s->components) if ((1 << i) & s->components)
ncomp++; ncomp++;
} }
if (s->thistogram) {
if (!s->width)
s->width = ctx->inputs[0]->w;
outlink->w = s->width * FFMAX(ncomp * (s->display_mode == 1), 1);
outlink->h = s->histogram_size * FFMAX(ncomp * (s->display_mode == 2), 1);
} else {
outlink->w = s->histogram_size * FFMAX(ncomp * (s->display_mode == 1), 1); outlink->w = s->histogram_size * FFMAX(ncomp * (s->display_mode == 1), 1);
outlink->h = (s->level_height + s->scale_height) * FFMAX(ncomp * (s->display_mode == 2), 1); outlink->h = (s->level_height + s->scale_height) * FFMAX(ncomp * (s->display_mode == 2), 1);
}
s->odesc = av_pix_fmt_desc_get(outlink->format); s->odesc = av_pix_fmt_desc_get(outlink->format);
s->dncomp = s->odesc->nb_components; s->dncomp = s->odesc->nb_components;
@ -251,16 +273,16 @@ static int filter_frame(AVFilterLink *inlink, AVFrame *in)
HistogramContext *s = inlink->dst->priv; HistogramContext *s = inlink->dst->priv;
AVFilterContext *ctx = inlink->dst; AVFilterContext *ctx = inlink->dst;
AVFilterLink *outlink = ctx->outputs[0]; AVFilterLink *outlink = ctx->outputs[0];
AVFrame *out; AVFrame *out = s->out;
int i, j, k, l, m; int i, j, k, l, m;
if (!s->thistogram || !out) {
out = ff_get_video_buffer(outlink, outlink->w, outlink->h); out = ff_get_video_buffer(outlink, outlink->w, outlink->h);
if (!out) { if (!out) {
av_frame_free(&in); av_frame_free(&in);
return AVERROR(ENOMEM); return AVERROR(ENOMEM);
} }
s->out = out;
out->pts = in->pts;
for (k = 0; k < 4 && out->data[k]; k++) { for (k = 0; k < 4 && out->data[k]; k++) {
const int is_chroma = (k == 1 || k == 2); const int is_chroma = (k == 1 || k == 2);
@ -282,19 +304,26 @@ static int filter_frame(AVFilterLink *inlink, AVFrame *in)
s->bg_color[k] * mult); s->bg_color[k] * mult);
} }
} }
}
for (m = 0, k = 0; k < s->ncomp; k++) { for (m = 0, k = 0; k < s->ncomp; k++) {
const int p = s->desc->comp[k].plane; const int p = s->desc->comp[k].plane;
const int max_value = s->histogram_size - 1 - s->start[p];
const int height = s->planeheight[p]; const int height = s->planeheight[p];
const int width = s->planewidth[p]; const int width = s->planewidth[p];
double max_hval_log; double max_hval_log;
unsigned max_hval = 0; unsigned max_hval = 0;
int start, startx; int starty, startx;
if (!((1 << k) & s->components)) if (!((1 << k) & s->components))
continue; continue;
if (s->thistogram) {
starty = m * s->histogram_size * (s->display_mode == 2);
startx = m++ * s->width * (s->display_mode == 1);
} else {
startx = m * s->histogram_size * (s->display_mode == 1); startx = m * s->histogram_size * (s->display_mode == 1);
start = m++ * (s->level_height + s->scale_height) * (s->display_mode == 2); starty = m++ * (s->level_height + s->scale_height) * (s->display_mode == 2);
}
if (s->histogram_size <= 256) { if (s->histogram_size <= 256) {
for (i = 0; i < height; i++) { for (i = 0; i < height; i++) {
@ -314,6 +343,23 @@ static int filter_frame(AVFilterLink *inlink, AVFrame *in)
max_hval = FFMAX(max_hval, s->histogram[i]); max_hval = FFMAX(max_hval, s->histogram[i]);
max_hval_log = log2(max_hval + 1); max_hval_log = log2(max_hval + 1);
if (s->thistogram) {
for (int i = 0; i < s->histogram_size; i++) {
int idx = s->histogram_size - i - 1;
int value = s->start[p];
if (s->levels_mode)
value += lrint(max_value * (log2(s->histogram[idx] + 1) / max_hval_log));
else
value += lrint(max_value * s->histogram[idx] / (float)max_hval);
if (s->histogram_size <= 256) {
s->out->data[p][(i + starty) * s->out->linesize[p] + startx + s->x_pos] = value;
} else {
AV_WN16(s->out->data[p] + (i + starty) * s->out->linesize[p] + startx * 2 + s->x_pos * 2, value);
}
}
} else {
for (i = 0; i < s->histogram_size; i++) { for (i = 0; i < s->histogram_size; i++) {
int col_height; int col_height;
@ -326,33 +372,46 @@ static int filter_frame(AVFilterLink *inlink, AVFrame *in)
for (j = s->level_height - 1; j >= col_height; j--) { for (j = s->level_height - 1; j >= col_height; j--) {
if (s->display_mode) { if (s->display_mode) {
for (l = 0; l < s->dncomp; l++) for (l = 0; l < s->dncomp; l++)
out->data[l][(j + start) * out->linesize[l] + startx + i] = s->fg_color[l]; out->data[l][(j + starty) * out->linesize[l] + startx + i] = s->fg_color[l];
} else { } else {
out->data[p][(j + start) * out->linesize[p] + startx + i] = 255; out->data[p][(j + starty) * out->linesize[p] + startx + i] = 255;
} }
} }
for (j = s->level_height + s->scale_height - 1; j >= s->level_height; j--) for (j = s->level_height + s->scale_height - 1; j >= s->level_height; j--)
out->data[p][(j + start) * out->linesize[p] + startx + i] = i; out->data[p][(j + starty) * out->linesize[p] + startx + i] = i;
} else { } else {
const int mult = s->mult; const int mult = s->mult;
for (j = s->level_height - 1; j >= col_height; j--) { for (j = s->level_height - 1; j >= col_height; j--) {
if (s->display_mode) { if (s->display_mode) {
for (l = 0; l < s->dncomp; l++) for (l = 0; l < s->dncomp; l++)
AV_WN16(out->data[l] + (j + start) * out->linesize[l] + startx * 2 + i * 2, s->fg_color[l] * mult); AV_WN16(out->data[l] + (j + starty) * out->linesize[l] + startx * 2 + i * 2, s->fg_color[l] * mult);
} else { } else {
AV_WN16(out->data[p] + (j + start) * out->linesize[p] + startx * 2 + i * 2, 255 * mult); AV_WN16(out->data[p] + (j + starty) * out->linesize[p] + startx * 2 + i * 2, 255 * mult);
} }
} }
for (j = s->level_height + s->scale_height - 1; j >= s->level_height; j--) for (j = s->level_height + s->scale_height - 1; j >= s->level_height; j--)
AV_WN16(out->data[p] + (j + start) * out->linesize[p] + startx * 2 + i * 2, i); AV_WN16(out->data[p] + (j + starty) * out->linesize[p] + startx * 2 + i * 2, i);
}
} }
} }
memset(s->histogram, 0, s->histogram_size * sizeof(unsigned)); memset(s->histogram, 0, s->histogram_size * sizeof(unsigned));
} }
out->pts = in->pts;
av_frame_free(&in); av_frame_free(&in);
s->x_pos++;
if (s->x_pos >= s->width)
s->x_pos = 0;
if (s->thistogram) {
AVFrame *clone = av_frame_clone(out);
if (!clone)
return AVERROR(ENOMEM);
return ff_filter_frame(outlink, clone);
}
return ff_filter_frame(outlink, out); return ff_filter_frame(outlink, out);
} }
@ -375,6 +434,8 @@ static const AVFilterPad outputs[] = {
{ NULL } { NULL }
}; };
#if CONFIG_HISTOGRAM_FILTER
AVFilter ff_vf_histogram = { AVFilter ff_vf_histogram = {
.name = "histogram", .name = "histogram",
.description = NULL_IF_CONFIG_SMALL("Compute and draw a histogram."), .description = NULL_IF_CONFIG_SMALL("Compute and draw a histogram."),
@ -384,3 +445,30 @@ AVFilter ff_vf_histogram = {
.outputs = outputs, .outputs = outputs,
.priv_class = &histogram_class, .priv_class = &histogram_class,
}; };
#endif /* CONFIG_HISTOGRAM_FILTER */
#if CONFIG_THISTOGRAM_FILTER
static const AVOption thistogram_options[] = {
{ "width", "set width", OFFSET(width), AV_OPT_TYPE_INT, {.i64=0}, 0, 8192, FLAGS},
{ "w", "set width", OFFSET(width), AV_OPT_TYPE_INT, {.i64=0}, 0, 8192, FLAGS},
COMMON_OPTIONS
{ "bgopacity", "set background opacity", OFFSET(bgopacity), AV_OPT_TYPE_FLOAT, {.dbl=0.5}, 0, 1, FLAGS},
{ "b", "set background opacity", OFFSET(bgopacity), AV_OPT_TYPE_FLOAT, {.dbl=0.5}, 0, 1, FLAGS},
{ NULL }
};
AVFILTER_DEFINE_CLASS(thistogram);
AVFilter ff_vf_thistogram = {
.name = "thistogram",
.description = NULL_IF_CONFIG_SMALL("Compute and draw a temporal histogram."),
.priv_size = sizeof(HistogramContext),
.query_formats = query_formats,
.inputs = inputs,
.outputs = outputs,
.priv_class = &thistogram_class,
};
#endif /* CONFIG_THISTOGRAM_FILTER */