[FFmpeg-devel] [PATCH] avfilter/vf_yaepblur: add yaepblur filter
leozhang
leozhang at qiyi.com
Mon Nov 25 11:53:05 EET 2019
Signed-off-by: leozhang <leozhang at qiyi.com>
---
This filter blur the input while preserving edges, with slice threads speed up.
My test speed is about 100fps on 1080p video with 16 threads, on my test machine whose cpu is E5-2660 v4 2.0GHz using 16 threads.
I gauss that i7-9700K 3.6GHz can run faster more.
The test command is
ffmpeg -s 1920x1080 -r 30 -i your_test_file.yuv -filter_threads 16 -vf yaepblur -f null -
doc/filters.texi | 22 +++
libavfilter/Makefile | 1 +
libavfilter/allfilters.c | 1 +
libavfilter/vf_yaepblur.c | 349 ++++++++++++++++++++++++++++++++++++++++++++++
4 files changed, 373 insertions(+)
create mode 100644 libavfilter/vf_yaepblur.c
diff --git a/doc/filters.texi b/doc/filters.texi
index c04421b..61e93d5 100644
--- a/doc/filters.texi
+++ b/doc/filters.texi
@@ -19775,6 +19775,28 @@ Only deinterlace frames marked as interlaced.
The default value is @code{all}.
@end table
+ at section yaepblur
+
+Apply blur filter while preserving edges ("yaepblur" means "yet another edge preserving blur filter").
+The algorithm is described in
+"J. S. Lee, Digital image enhancement and noise filtering by use of local statistics, IEEE Trans. Pattern Anal. Mach. Intell. PAMI-2, 1980."
+
+It accepts the following parameters:
+
+ at table @option
+ at item radius, r
+Set the window radius. Default value is 3.
+
+ at item planes, p
+Set which planes to filter. Default is only the first plane.
+
+ at item sigma, s
+Set blur strength. Default value is 128.
+ at end table
+
+ at subsection Commands
+This filter supports same @ref{commands} as options.
+
@section zoompan
Apply Zoom & Pan effect.
diff --git a/libavfilter/Makefile b/libavfilter/Makefile
index 6838d5c..b490a44 100644
--- a/libavfilter/Makefile
+++ b/libavfilter/Makefile
@@ -442,6 +442,7 @@ OBJS-$(CONFIG_XSTACK_FILTER) += vf_stack.o framesync.o
OBJS-$(CONFIG_YADIF_FILTER) += vf_yadif.o yadif_common.o
OBJS-$(CONFIG_YADIF_CUDA_FILTER) += vf_yadif_cuda.o vf_yadif_cuda.ptx.o \
yadif_common.o
+OBJS-$(CONFIG_YAEPBLUR_FILTER) += vf_yaepblur.o
OBJS-$(CONFIG_ZMQ_FILTER) += f_zmq.o
OBJS-$(CONFIG_ZOOMPAN_FILTER) += vf_zoompan.o
OBJS-$(CONFIG_ZSCALE_FILTER) += vf_zscale.o
diff --git a/libavfilter/allfilters.c b/libavfilter/allfilters.c
index 7c1e19e..8f41186 100644
--- a/libavfilter/allfilters.c
+++ b/libavfilter/allfilters.c
@@ -420,6 +420,7 @@ extern AVFilter ff_vf_xmedian;
extern AVFilter ff_vf_xstack;
extern AVFilter ff_vf_yadif;
extern AVFilter ff_vf_yadif_cuda;
+extern AVFilter ff_vf_yaepblur;
extern AVFilter ff_vf_zmq;
extern AVFilter ff_vf_zoompan;
extern AVFilter ff_vf_zscale;
diff --git a/libavfilter/vf_yaepblur.c b/libavfilter/vf_yaepblur.c
new file mode 100644
index 0000000..ef6fbc9
--- /dev/null
+++ b/libavfilter/vf_yaepblur.c
@@ -0,0 +1,349 @@
+/*
+ * Copyright (C) 2019 Leo Zhang <leozhang at qiyi.com>
+
+ * 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
+ * yaep(yet another edge preserving) blur filter
+ *
+ * This implementation is based on an algorithm described in
+ * "J. S. Lee, Digital image enhancement and noise filtering by use of local statistics, IEEE Trans. Pattern
+ * Anal. Mach. Intell. PAMI-2, 1980."
+ */
+
+#include "libavutil/opt.h"
+#include "libavutil/imgutils.h"
+#include "avfilter.h"
+#include "internal.h"
+
+typedef struct YAEPContext {
+ const AVClass *class;
+
+ int planes;
+ int radius;
+ int sigma;
+
+ int nb_planes;
+ int planewidth[4];
+ int planeheight[4];
+ int depth;
+
+ uint64_t *sat; ///< summed area table
+ uint64_t *square_sat; ///< square summed area table
+ int sat_linesize;
+
+ int (*pre_calculate_row)(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs);
+ int (*filter_slice )(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs);
+} YAEPContext;
+
+static av_cold void uninit(AVFilterContext *ctx)
+{
+ YAEPContext *s = ctx->priv;
+ av_freep(&s->sat);
+ av_freep(&s->square_sat);
+}
+
+static int query_formats(AVFilterContext *ctx)
+{
+ static const enum AVPixelFormat pix_fmts[] = {
+ AV_PIX_FMT_YUVA444P, AV_PIX_FMT_YUV444P, AV_PIX_FMT_YUV440P,
+ AV_PIX_FMT_YUVJ444P, AV_PIX_FMT_YUVJ440P,
+ AV_PIX_FMT_YUVA422P, AV_PIX_FMT_YUV422P, AV_PIX_FMT_YUVA420P, AV_PIX_FMT_YUV420P,
+ AV_PIX_FMT_YUVJ422P, AV_PIX_FMT_YUVJ420P,
+ AV_PIX_FMT_YUVJ411P, AV_PIX_FMT_YUV411P, AV_PIX_FMT_YUV410P,
+ AV_PIX_FMT_YUV420P9, AV_PIX_FMT_YUV422P9, AV_PIX_FMT_YUV444P9,
+ AV_PIX_FMT_YUV420P10, AV_PIX_FMT_YUV422P10, AV_PIX_FMT_YUV444P10,
+ AV_PIX_FMT_YUV420P12, AV_PIX_FMT_YUV422P12, AV_PIX_FMT_YUV444P12, AV_PIX_FMT_YUV440P12,
+ AV_PIX_FMT_YUV420P14, AV_PIX_FMT_YUV422P14, AV_PIX_FMT_YUV444P14,
+ AV_PIX_FMT_YUV420P16, AV_PIX_FMT_YUV422P16, AV_PIX_FMT_YUV444P16,
+ AV_PIX_FMT_YUVA420P9, AV_PIX_FMT_YUVA422P9, AV_PIX_FMT_YUVA444P9,
+ AV_PIX_FMT_YUVA420P10, AV_PIX_FMT_YUVA422P10, AV_PIX_FMT_YUVA444P10,
+ AV_PIX_FMT_YUVA422P12, AV_PIX_FMT_YUVA444P12,
+ AV_PIX_FMT_YUVA420P16, AV_PIX_FMT_YUVA422P16, AV_PIX_FMT_YUVA444P16,
+ 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_GRAY8, AV_PIX_FMT_GRAY9, AV_PIX_FMT_GRAY10, AV_PIX_FMT_GRAY12, AV_PIX_FMT_GRAY14, AV_PIX_FMT_GRAY16,
+ AV_PIX_FMT_NONE
+ };
+
+ return ff_set_common_formats(ctx, ff_make_format_list(pix_fmts));
+}
+
+typedef struct ThreadData {
+ int width;
+ int height;
+ int src_linesize;
+ int dst_linesize;
+ uint8_t *src;
+ uint8_t *dst;
+} ThreadData;
+
+#define PRE_CALCULATE_ROW(type, name) \
+static int pre_calculate_row_##name(AVFilterContext *ctx, void *arg, \
+ int jobnr, int nb_jobs) \
+{ \
+ ThreadData *td = arg; \
+ YAEPContext *s = ctx->priv; \
+ \
+ const int width = td->width; \
+ const int height = td->height; \
+ const int linesize = td->src_linesize / sizeof(type); \
+ const int sat_linesize = s->sat_linesize; \
+ \
+ const int starty = height * jobnr / nb_jobs; \
+ const int endy = height * (jobnr+1) / nb_jobs; \
+ \
+ uint64_t *sat = s->sat + (starty + 1) * sat_linesize; \
+ uint64_t *square_sat = s->square_sat + (starty + 1) * sat_linesize; \
+ const type *src = (const type *)td->src + starty * linesize; \
+ \
+ int x, y; \
+ \
+ for (y = starty; y < endy; y++) { \
+ for (x = 0; x < width; x++) { \
+ sat[x+1] = sat[x] + src[x]; \
+ square_sat[x+1] = square_sat[x] + (uint64_t)src[x] * src[x]; \
+ } \
+ sat += sat_linesize; \
+ square_sat += sat_linesize; \
+ src += linesize; \
+ } \
+ \
+ return 0; \
+}
+
+PRE_CALCULATE_ROW(uint8_t, byte)
+PRE_CALCULATE_ROW(uint16_t, word)
+
+static int pre_calculate_col(AVFilterContext *ctx, void *arg,
+ int jobnr, int nb_jobs)
+{
+ ThreadData *td = arg;
+ YAEPContext *s = ctx->priv;
+
+ const int width = td->width;
+ const int height = td->height;
+ const int sat_linesize = s->sat_linesize;
+
+ const int startx = width * jobnr / nb_jobs;
+ const int endx = width * (jobnr + 1) / nb_jobs;
+
+ uint64_t *sat, *square_sat;
+ int x, y;
+
+ for (x = startx; x < endx; x++) {
+ sat = s->sat + x + 1;
+ square_sat = s->square_sat + x + 1;
+ for (y = 0; y < height; y++) {
+ *(sat+sat_linesize) += *sat;
+ *(square_sat+sat_linesize) += *square_sat;
+ sat += sat_linesize;
+ square_sat += sat_linesize;
+ }
+ }
+
+ return 0;
+}
+
+#define FILTER_SLICE(type, name) \
+static int filter_slice_##name(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs) \
+{ \
+ ThreadData *td = arg; \
+ YAEPContext *s = ctx->priv; \
+ \
+ const int width = td->width; \
+ const int height = td->height; \
+ const int src_linesize = td->src_linesize / sizeof(type); \
+ const int dst_linesize = td->dst_linesize / sizeof(type); \
+ const int sat_linesize = s->sat_linesize; \
+ const int sigma = s->sigma; \
+ const int radius = s->radius; \
+ \
+ uint64_t *sat = s->sat; \
+ uint64_t *square_sat = s->square_sat; \
+ const type *src = (const type *)td->src; \
+ type *dst = (type *)td->dst; \
+ \
+ const int starty = height * jobnr / nb_jobs; \
+ const int endy = height * (jobnr + 1) / nb_jobs; \
+ \
+ int x, y; \
+ int lower_x, higher_x; \
+ int lower_y, higher_y; \
+ int dist_y, count; \
+ uint64_t sum, square_sum, mean, var; \
+ \
+ for (y = starty; y < endy; y++) { \
+ lower_y = y - radius < 0 ? 0 : y - radius; \
+ higher_y = y + radius + 1 > height ? height : y + radius + 1; \
+ dist_y = higher_y - lower_y; \
+ for (x = 0; x < width; x++) { \
+ lower_x = x - radius < 0 ? 0 : x - radius; \
+ higher_x = x + radius + 1 > width ? width : x + radius + 1; \
+ count = dist_y * (higher_x - lower_x); \
+ sum = sat[higher_y * sat_linesize + higher_x] \
+ - sat[higher_y * sat_linesize + lower_x] \
+ - sat[lower_y * sat_linesize + higher_x] \
+ + sat[lower_y * sat_linesize + lower_x]; \
+ square_sum = square_sat[higher_y * sat_linesize + higher_x] \
+ - square_sat[higher_y * sat_linesize + lower_x] \
+ - square_sat[lower_y * sat_linesize + higher_x] \
+ + square_sat[lower_y * sat_linesize + lower_x]; \
+ mean = sum / count; \
+ var = (square_sum - sum * sum / count) / count; \
+ dst[y * dst_linesize + x] = (sigma * mean + var * src[y * src_linesize + x]) / (sigma + var); \
+ } \
+ } \
+ return 0; \
+}
+
+FILTER_SLICE(uint8_t, byte)
+FILTER_SLICE(uint16_t, word)
+
+static int filter_frame(AVFilterLink *inlink, AVFrame *in)
+{
+ AVFilterContext *ctx = inlink->dst;
+ YAEPContext *s = ctx->priv;
+ AVFilterLink *outlink = ctx->outputs[0];
+ AVFrame *out;
+ int plane;
+ const int nb_threads = ff_filter_get_nb_threads(ctx);
+ ThreadData td;
+
+ if (av_frame_is_writable(in)) {
+ out = in;
+ } else {
+ out = ff_get_video_buffer(outlink, outlink->w, outlink->h);
+ if (!out) {
+ av_frame_free(&in);
+ return AVERROR(ENOMEM);
+ }
+ av_frame_copy_props(out, in);
+ }
+
+ for (plane = 0; plane < s->nb_planes; plane++) {
+ if (!s->radius || !(s->planes & (1<<plane))) {
+ if (out != in) {
+ av_image_copy_plane(out->data[plane], out->linesize[plane],
+ in->data[plane], in->linesize[plane],
+ s->planewidth[plane] * ((s->depth + 7) / 8),
+ s->planeheight[plane]);
+ }
+ continue;
+ }
+
+ td.width = s->planewidth[plane];
+ td.height = s->planeheight[plane];
+ td.src = in->data[plane];
+ td.src_linesize = in->linesize[plane];
+ ctx->internal->execute(ctx, s->pre_calculate_row, &td, NULL, FFMIN(td.height, nb_threads));
+ ctx->internal->execute(ctx, pre_calculate_col, &td, NULL, FFMIN(td.width, nb_threads));
+
+ td.dst = out->data[plane];
+ td.dst_linesize = out->linesize[plane];
+ ctx->internal->execute(ctx, s->filter_slice, &td, NULL, FFMIN(td.height, nb_threads));
+ }
+
+ if (out != in)
+ av_frame_free(&in);
+
+ return ff_filter_frame(outlink, out);
+}
+
+static int config_input(AVFilterLink *inlink)
+{
+ YAEPContext *s = inlink->dst->priv;
+ const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(inlink->format);
+
+ s->depth = desc->comp[0].depth;
+ s->planewidth[1] = s->planewidth[2] = AV_CEIL_RSHIFT(inlink->w, desc->log2_chroma_w);
+ s->planewidth[0] = s->planewidth[3] = inlink->w;
+ s->planeheight[1] = s->planeheight[2] = AV_CEIL_RSHIFT(inlink->h, desc->log2_chroma_h);
+ s->planeheight[0] = s->planeheight[3] = inlink->h;
+ s->nb_planes = av_pix_fmt_count_planes(inlink->format);
+
+ s->radius = FFMIN(s->radius, AV_CEIL_RSHIFT(FFMIN(inlink->w, inlink->h), 1));
+
+ if (s->depth <= 8) {
+ s->pre_calculate_row = pre_calculate_row_byte;
+ s->filter_slice = filter_slice_byte;
+ } else {
+ s->pre_calculate_row = pre_calculate_row_word;
+ s->filter_slice = filter_slice_word;
+ }
+
+ // padding one row on the top, and padding one col on the left, that is why + 1 below
+ s->sat_linesize = inlink->w + 1;
+ s->sat = av_mallocz_array(inlink->h + 1, s->sat_linesize*sizeof(*s->sat));
+ if (!s->sat)
+ return AVERROR(ENOMEM);
+
+ s->square_sat = av_mallocz_array(inlink->h + 1, s->sat_linesize*sizeof(*s->square_sat));
+ if (!s->square_sat)
+ return AVERROR(ENOMEM);
+
+ return 0;
+}
+
+static const AVFilterPad yaep_inputs[] = {
+ {
+ .name = "default",
+ .type = AVMEDIA_TYPE_VIDEO,
+ .config_props = config_input,
+ .filter_frame = filter_frame,
+ },
+ { NULL }
+};
+
+static const AVFilterPad yaep_outputs[] = {
+ {
+ .name = "default",
+ .type = AVMEDIA_TYPE_VIDEO,
+ },
+ { NULL }
+};
+
+#define OFFSET(x) offsetof(YAEPContext, x)
+#define FLAGS AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_VIDEO_PARAM|AV_OPT_FLAG_RUNTIME_PARAM
+
+static const AVOption yaep_options[] = {
+ { "radius", "set window radius", OFFSET(radius), AV_OPT_TYPE_INT, {.i64=3}, 0, INT_MAX, .flags=FLAGS },
+ { "r" , "set window radius", OFFSET(radius), AV_OPT_TYPE_INT, {.i64=3}, 0, INT_MAX, .flags=FLAGS },
+ { "planes", "set planes to filter", OFFSET(planes), AV_OPT_TYPE_INT, {.i64=1}, 0, 0xF, .flags=FLAGS },
+ { "p", "set planes to filter", OFFSET(planes), AV_OPT_TYPE_INT, {.i64=1}, 0, 0xF, .flags=FLAGS },
+ { "sigma", "set blur strength", OFFSET(sigma), AV_OPT_TYPE_INT, {.i64=128}, 1, INT_MAX, .flags=FLAGS },
+ { "s", "set blur strength", OFFSET(sigma), AV_OPT_TYPE_INT, {.i64=128}, 1, INT_MAX, .flags=FLAGS },
+ { NULL }
+};
+
+AVFILTER_DEFINE_CLASS(yaep);
+
+AVFilter ff_vf_yaepblur = {
+ .name = "yaepblur",
+ .description = NULL_IF_CONFIG_SMALL("Yet another edge preserving blur filter."),
+ .priv_size = sizeof(YAEPContext),
+ .priv_class = &yaep_class,
+ .uninit = uninit,
+ .query_formats = query_formats,
+ .inputs = yaep_inputs,
+ .outputs = yaep_outputs,
+ .flags = AVFILTER_FLAG_SUPPORT_TIMELINE_GENERIC | AVFILTER_FLAG_SLICE_THREADS,
+ .process_command = ff_filter_process_command,
+};
--
1.8.3.1
More information about the ffmpeg-devel
mailing list