From: Paul B Mahol <onemda@gmail.com>
To: FFmpeg development discussions and patches <ffmpeg-devel@ffmpeg.org>
Subject: [FFmpeg-devel] [PATCH] avfilter: add corr video filter
Date: Tue, 6 Dec 2022 22:29:24 +0100
Message-ID: <CAPYw7P7--OqQkZjnN-w9U=+TdGtBGGEUkDcxq4EXSBCd29Wbng@mail.gmail.com> (raw)
[-- Attachment #1: Type: text/plain, Size: 49 bytes --]
Adds yet another metric filter.
Patch attached.
[-- Attachment #2: 0001-avfilter-add-corr-video-filter.patch --]
[-- Type: text/x-patch, Size: 15825 bytes --]
From 47f30e54b184e4b96f2e9da78671ca5d85c83762 Mon Sep 17 00:00:00 2001
From: Paul B Mahol <onemda@gmail.com>
Date: Tue, 6 Dec 2022 21:31:33 +0100
Subject: [PATCH] avfilter: add corr video filter
Signed-off-by: Paul B Mahol <onemda@gmail.com>
---
doc/filters.texi | 22 +++
libavfilter/Makefile | 1 +
libavfilter/allfilters.c | 1 +
libavfilter/vf_corr.c | 339 +++++++++++++++++++++++++++++++++++++++
4 files changed, 363 insertions(+)
create mode 100644 libavfilter/vf_corr.c
diff --git a/doc/filters.texi b/doc/filters.texi
index 4b74bce34f..252741b254 100644
--- a/doc/filters.texi
+++ b/doc/filters.texi
@@ -10097,6 +10097,28 @@ ffmpeg -f lavfi -i nullsrc=s=100x100,coreimage=filter=CIQRCodeGenerator@@inputMe
@end example
@end itemize
+@section corr
+
+Obtain the correlation between two input videos.
+
+This filter takes two input videos.
+
+Both input videos must have the same resolution and pixel format for
+this filter to work correctly. Also it assumes that both inputs
+have the same number of frames, which are compared one by one.
+
+The obtained per component, average, min and max correlation is printed through
+the logging system.
+
+The filter stores the calculated correlation of each frame in frame metadata.
+
+In the below example the input file @file{main.mpg} being processed is compared
+with the reference file @file{ref.mpg}.
+
+@example
+ffmpeg -i main.mpg -i ref.mpg -lavfi corr -f null -
+@end example
+
@section cover_rect
Cover a rectangular object
diff --git a/libavfilter/Makefile b/libavfilter/Makefile
index 2791b6a950..9a850d5dc5 100644
--- a/libavfilter/Makefile
+++ b/libavfilter/Makefile
@@ -243,6 +243,7 @@ OBJS-$(CONFIG_CONVOLUTION_OPENCL_FILTER) += vf_convolution_opencl.o opencl.o
OBJS-$(CONFIG_CONVOLVE_FILTER) += vf_convolve.o framesync.o
OBJS-$(CONFIG_COPY_FILTER) += vf_copy.o
OBJS-$(CONFIG_COREIMAGE_FILTER) += vf_coreimage.o
+OBJS-$(CONFIG_CORR_FILTER) += vf_corr.o framesync.o
OBJS-$(CONFIG_COVER_RECT_FILTER) += vf_cover_rect.o lavfutils.o
OBJS-$(CONFIG_CROP_FILTER) += vf_crop.o
OBJS-$(CONFIG_CROPDETECT_FILTER) += vf_cropdetect.o
diff --git a/libavfilter/allfilters.c b/libavfilter/allfilters.c
index 3ff20e76ce..2ece5c15c8 100644
--- a/libavfilter/allfilters.c
+++ b/libavfilter/allfilters.c
@@ -223,6 +223,7 @@ extern const AVFilter ff_vf_convolution_opencl;
extern const AVFilter ff_vf_convolve;
extern const AVFilter ff_vf_copy;
extern const AVFilter ff_vf_coreimage;
+extern const AVFilter ff_vf_corr;
extern const AVFilter ff_vf_cover_rect;
extern const AVFilter ff_vf_crop;
extern const AVFilter ff_vf_cropdetect;
diff --git a/libavfilter/vf_corr.c b/libavfilter/vf_corr.c
new file mode 100644
index 0000000000..3e3951d001
--- /dev/null
+++ b/libavfilter/vf_corr.c
@@ -0,0 +1,339 @@
+/*
+ * This file is part of FFmpeg.
+ *
+ * FFmpeg is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2.1 of the License, or (at your option) any later version.
+ *
+ * FFmpeg is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
+ * Lesser General Public License for more details.
+ *
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with FFmpeg; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
+ */
+
+/**
+ * @file
+ * Calculate the correlation between two input videos.
+ */
+
+#include "libavutil/avstring.h"
+#include "libavutil/opt.h"
+#include "libavutil/pixdesc.h"
+#include "avfilter.h"
+#include "drawutils.h"
+#include "formats.h"
+#include "framesync.h"
+#include "internal.h"
+#include "video.h"
+
+typedef struct CorrContext {
+ const AVClass *class;
+ FFFrameSync fs;
+ double score, min_score, max_score, score_comp[4];
+ uint64_t nb_frames;
+ int is_rgb;
+ uint8_t rgba_map[4];
+ int max[4];
+ char comps[4];
+ int nb_components;
+ int planewidth[4];
+ int planeheight[4];
+ int (*filter_slice)(AVFilterContext *ctx, void *arg,
+ int jobnr, int nb_jobs);
+} CorrContext;
+
+#define OFFSET(x) offsetof(CorrContext, x)
+#define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM
+
+static void set_meta(AVFilterContext *ctx,
+ AVDictionary **metadata, const char *key, char comp, float d)
+{
+ char value[128];
+ snprintf(value, sizeof(value), "%f", d);
+ if (comp) {
+ char key2[128];
+ snprintf(key2, sizeof(key2), "lavfi.%s.%s%s%c",
+ ctx->filter->name, ctx->filter->name, key, comp);
+ av_dict_set(metadata, key2, value, 0);
+ } else {
+ char key2[128];
+ snprintf(key2, sizeof(key2), "lavfi.%s.%s%s",
+ ctx->filter->name, ctx->filter->name, key);
+ av_dict_set(metadata, key2, value, 0);
+ }
+}
+
+#define CORR(type, name) \
+static void f##name(AVFilterContext *ctx, AVFrame *master, \
+ AVFrame *ref, double *comp_score) \
+{ \
+ CorrContext *s = ctx->priv; \
+ \
+ for (int c = 0; c < s->nb_components; c++) { \
+ const ptrdiff_t linesize1 = master->linesize[c] / \
+ sizeof(type); \
+ const ptrdiff_t linesize2 = ref->linesize[c] / \
+ sizeof(type); \
+ const type *src1 = (const type *)master->data[c]; \
+ const type *src2 = (const type *)ref->data[c]; \
+ const int h = s->planeheight[c]; \
+ const int w = s->planewidth[c]; \
+ const float scale = 1.f / s->max[c]; \
+ uint64_t sum1 = 0, sum2 = 0; \
+ float sum12, sum1q, sum2q; \
+ float sumq, mean1, mean2; \
+ \
+ for (int y = 0; y < h; y++) { \
+ for (int x = 0; x < w; x++) { \
+ sum1 += src1[x]; \
+ sum2 += src2[x]; \
+ } \
+ \
+ src1 += linesize1; \
+ src2 += linesize2; \
+ } \
+ \
+ mean1 = scale * (sum1 /(double)(w * h)); \
+ mean2 = scale * (sum2 /(double)(w * h)); \
+ \
+ src1 = (const type *)master->data[c]; \
+ src2 = (const type *)ref->data[c]; \
+ \
+ sum12 = 0.f; \
+ sum1q = 0.f; \
+ sum2q = 0.f; \
+ \
+ for (int y = 0; y < h; y++) { \
+ for (int x = 0; x < w; x++) { \
+ const float f1 = scale * src1[x] - mean1; \
+ const float f2 = scale * src2[x] - mean2; \
+ \
+ sum12 += f1 * f2; \
+ sum1q += f1 * f1; \
+ sum2q += f2 * f2; \
+ } \
+ \
+ src1 += linesize1; \
+ src2 += linesize2; \
+ } \
+ \
+ sumq = sqrtf(sum1q * sum2q); \
+ if (sumq > 0.f) { \
+ comp_score[c] = av_clipf(sum12 / sumq,-1.f,1.f); \
+ } else { \
+ comp_score[c] = sum1q == sum2q ? 1.f : 0.f; \
+ } \
+ } \
+}
+
+CORR(uint8_t, corr8)
+CORR(uint16_t, corr16)
+
+static int do_corr(FFFrameSync *fs)
+{
+ AVFilterContext *ctx = fs->parent;
+ CorrContext *s = ctx->priv;
+ AVFrame *master, *ref;
+ double comp_score[4], score = 0.;
+ AVDictionary **metadata;
+ int ret;
+
+ ret = ff_framesync_dualinput_get(fs, &master, &ref);
+ if (ret < 0)
+ return ret;
+ if (ctx->is_disabled || !ref)
+ return ff_filter_frame(ctx->outputs[0], master);
+ metadata = &master->metadata;
+
+ if (s->max[0] > 255) {
+ fcorr16(ctx, master, ref, comp_score);
+ } else {
+ fcorr8(ctx, master, ref, comp_score);
+ }
+
+ for (int c = 0; c < s->nb_components; c++)
+ score += comp_score[c];
+ score /= s->nb_components;
+ s->score += score;
+
+ s->min_score = fmin(s->min_score, score);
+ s->max_score = fmax(s->max_score, score);
+
+ for (int c = 0; c < s->nb_components; c++)
+ s->score_comp[c] += comp_score[c];
+ s->nb_frames++;
+
+ for (int j = 0; j < s->nb_components; j++) {
+ int c = s->is_rgb ? s->rgba_map[j] : j;
+ set_meta(ctx, metadata, ".", s->comps[j], comp_score[c]);
+ }
+ set_meta(ctx, metadata, "_avg", 0, score);
+
+ return ff_filter_frame(ctx->outputs[0], master);
+}
+
+static av_cold int init(AVFilterContext *ctx)
+{
+ CorrContext *s = ctx->priv;
+
+ s->fs.on_event = do_corr;
+
+ return 0;
+}
+
+static const enum AVPixelFormat pix_fmts[] = {
+ AV_PIX_FMT_GRAY8, AV_PIX_FMT_GRAY9, AV_PIX_FMT_GRAY10, AV_PIX_FMT_GRAY12, AV_PIX_FMT_GRAY14, AV_PIX_FMT_GRAY16,
+#define PF_NOALPHA(suf) AV_PIX_FMT_YUV420##suf, AV_PIX_FMT_YUV422##suf, AV_PIX_FMT_YUV444##suf
+#define PF_ALPHA(suf) AV_PIX_FMT_YUVA420##suf, AV_PIX_FMT_YUVA422##suf, AV_PIX_FMT_YUVA444##suf
+#define PF(suf) PF_NOALPHA(suf), PF_ALPHA(suf)
+ PF(P), PF(P9), PF(P10), PF_NOALPHA(P12), PF_NOALPHA(P14), PF(P16),
+ AV_PIX_FMT_YUV440P, AV_PIX_FMT_YUV411P, AV_PIX_FMT_YUV410P,
+ AV_PIX_FMT_YUVJ411P, AV_PIX_FMT_YUVJ420P, AV_PIX_FMT_YUVJ422P,
+ AV_PIX_FMT_YUVJ440P, AV_PIX_FMT_YUVJ444P,
+ AV_PIX_FMT_GBRP, AV_PIX_FMT_GBRP9, AV_PIX_FMT_GBRP10,
+ AV_PIX_FMT_GBRP12, AV_PIX_FMT_GBRP14, AV_PIX_FMT_GBRP16,
+ AV_PIX_FMT_GBRAP, AV_PIX_FMT_GBRAP10, AV_PIX_FMT_GBRAP12, AV_PIX_FMT_GBRAP16,
+ AV_PIX_FMT_NONE
+};
+
+static int config_input_ref(AVFilterLink *inlink)
+{
+ const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(inlink->format);
+ AVFilterContext *ctx = inlink->dst;
+ CorrContext *s = ctx->priv;
+
+ s->nb_components = desc->nb_components;
+ if (ctx->inputs[0]->w != ctx->inputs[1]->w ||
+ ctx->inputs[0]->h != ctx->inputs[1]->h) {
+ av_log(ctx, AV_LOG_ERROR, "Width and height of input videos must be same.\n");
+ return AVERROR(EINVAL);
+ }
+
+ s->is_rgb = ff_fill_rgba_map(s->rgba_map, inlink->format) >= 0;
+ s->comps[0] = s->is_rgb ? 'R' : 'Y' ;
+ s->comps[1] = s->is_rgb ? 'G' : 'U' ;
+ s->comps[2] = s->is_rgb ? 'B' : 'V' ;
+ s->comps[3] = 'A';
+
+ s->planeheight[1] = s->planeheight[2] = AV_CEIL_RSHIFT(inlink->h, desc->log2_chroma_h);
+ s->planeheight[0] = s->planeheight[3] = inlink->h;
+ s->planewidth[1] = s->planewidth[2] = AV_CEIL_RSHIFT(inlink->w, desc->log2_chroma_w);
+ s->planewidth[0] = s->planewidth[3] = inlink->w;
+
+ s->min_score = +INFINITY;
+ s->max_score = -INFINITY;
+
+ s->max[0] = (1 << desc->comp[0].depth) - 1;
+ s->max[1] = (1 << desc->comp[1].depth) - 1;
+ s->max[2] = (1 << desc->comp[2].depth) - 1;
+ s->max[3] = (1 << desc->comp[3].depth) - 1;
+
+ return 0;
+}
+
+static int config_output(AVFilterLink *outlink)
+{
+ AVFilterContext *ctx = outlink->src;
+ CorrContext *s = ctx->priv;
+ AVFilterLink *mainlink = ctx->inputs[0];
+ int ret;
+
+ ret = ff_framesync_init_dualinput(&s->fs, ctx);
+ if (ret < 0)
+ return ret;
+ outlink->w = mainlink->w;
+ outlink->h = mainlink->h;
+ outlink->time_base = mainlink->time_base;
+ outlink->sample_aspect_ratio = mainlink->sample_aspect_ratio;
+ outlink->frame_rate = mainlink->frame_rate;
+ if ((ret = ff_framesync_configure(&s->fs)) < 0)
+ return ret;
+
+ outlink->time_base = s->fs.time_base;
+
+ if (av_cmp_q(mainlink->time_base, outlink->time_base) ||
+ av_cmp_q(ctx->inputs[1]->time_base, outlink->time_base))
+ av_log(ctx, AV_LOG_WARNING, "not matching timebases found between first input: %d/%d and second input %d/%d, results may be incorrect!\n",
+ mainlink->time_base.num, mainlink->time_base.den,
+ ctx->inputs[1]->time_base.num, ctx->inputs[1]->time_base.den);
+
+ return 0;
+}
+
+static int activate(AVFilterContext *ctx)
+{
+ CorrContext *s = ctx->priv;
+ return ff_framesync_activate(&s->fs);
+}
+
+static av_cold void uninit(AVFilterContext *ctx)
+{
+ CorrContext *s = ctx->priv;
+
+ if (s->nb_frames > 0) {
+ char buf[256];
+
+ buf[0] = 0;
+ for (int j = 0; j < s->nb_components; j++) {
+ int c = s->is_rgb ? s->rgba_map[j] : j;
+ av_strlcatf(buf, sizeof(buf), " %c:%f", s->comps[j], s->score_comp[c] / s->nb_frames);
+ }
+
+ av_log(ctx, AV_LOG_INFO, "%s%s average:%f min:%f max:%f\n",
+ ctx->filter->name,
+ buf,
+ s->score / s->nb_frames,
+ s->min_score,
+ s->max_score);
+ }
+
+ ff_framesync_uninit(&s->fs);
+}
+
+static const AVFilterPad corr_inputs[] = {
+ {
+ .name = "main",
+ .type = AVMEDIA_TYPE_VIDEO,
+ },{
+ .name = "reference",
+ .type = AVMEDIA_TYPE_VIDEO,
+ .config_props = config_input_ref,
+ },
+};
+
+static const AVFilterPad corr_outputs[] = {
+ {
+ .name = "default",
+ .type = AVMEDIA_TYPE_VIDEO,
+ .config_props = config_output,
+ },
+};
+
+static const AVOption options[] = {
+ { NULL }
+};
+
+#define corr_options options
+FRAMESYNC_DEFINE_CLASS(corr, CorrContext, fs);
+
+const AVFilter ff_vf_corr = {
+ .name = "corr",
+ .description = NULL_IF_CONFIG_SMALL("Calculate the correlation between two video streams."),
+ .preinit = corr_framesync_preinit,
+ .init = init,
+ .uninit = uninit,
+ .activate = activate,
+ .priv_size = sizeof(CorrContext),
+ .priv_class = &corr_class,
+ FILTER_INPUTS(corr_inputs),
+ FILTER_OUTPUTS(corr_outputs),
+ FILTER_PIXFMTS_ARRAY(pix_fmts),
+ .flags = AVFILTER_FLAG_SUPPORT_TIMELINE_INTERNAL |
+ AVFILTER_FLAG_SLICE_THREADS |
+ AVFILTER_FLAG_METADATA_ONLY,
+};
--
2.37.2
[-- Attachment #3: Type: text/plain, Size: 251 bytes --]
_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".
next reply other threads:[~2022-12-06 21:29 UTC|newest]
Thread overview: 2+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-12-06 21:29 Paul B Mahol [this message]
2022-12-07 10:30 ` Paul B Mahol
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to='CAPYw7P7--OqQkZjnN-w9U=+TdGtBGGEUkDcxq4EXSBCd29Wbng@mail.gmail.com' \
--to=onemda@gmail.com \
--cc=ffmpeg-devel@ffmpeg.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Git Inbox Mirror of the ffmpeg-devel mailing list - see https://ffmpeg.org/mailman/listinfo/ffmpeg-devel
This inbox may be cloned and mirrored by anyone:
git clone --mirror https://master.gitmailbox.com/ffmpegdev/0 ffmpegdev/git/0.git
# If you have public-inbox 1.1+ installed, you may
# initialize and index your mirror using the following commands:
public-inbox-init -V2 ffmpegdev ffmpegdev/ https://master.gitmailbox.com/ffmpegdev \
ffmpegdev@gitmailbox.com
public-inbox-index ffmpegdev
Example config snippet for mirrors.
AGPL code for this site: git clone https://public-inbox.org/public-inbox.git