diff mbox series

[FFmpeg-devel] avfilter: add iir video filter

Message ID 20200525145831.26292-1-onemda@gmail.com
State New
Headers show
Series [FFmpeg-devel] avfilter: add iir video filter
Related show

Checks

Context Check Description
andriy/default pending
andriy/make success Make finished
andriy/make_fate success Make fate finished

Commit Message

Paul B Mahol May 25, 2020, 2:58 p.m. UTC
Signed-off-by: Paul B Mahol <onemda@gmail.com>
---
 doc/filters.texi         |  18 ++
 libavfilter/Makefile     |   1 +
 libavfilter/allfilters.c |   1 +
 libavfilter/vf_iir.c     | 410 +++++++++++++++++++++++++++++++++++++++
 4 files changed, 430 insertions(+)
 create mode 100644 libavfilter/vf_iir.c
diff mbox series

Patch

diff --git a/doc/filters.texi b/doc/filters.texi
index 773473f721..7675a0e9b1 100644
--- a/doc/filters.texi
+++ b/doc/filters.texi
@@ -12500,6 +12500,24 @@  further computations. This allows inserting the idet filter as a low computation
 method to clean up the interlaced flag
 @end table
 
+@section iir
+Apply 2D Infinite Impulse Response filter to input video stream.
+
+The filter accepts the following options:
+
+@table @option
+@item hn
+@item vn
+Set horizontal and vertical numerator coefficients of transfer function.
+
+@item hd
+@item vd
+Set horizontal and vertical denominator coefficients of transfer function.
+
+@item planes
+Set which planes to filter. Default is all. Allowed range is from 0 to 15.
+@end table
+
 @section il
 
 Deinterleave or interleave fields.
diff --git a/libavfilter/Makefile b/libavfilter/Makefile
index 994a4172a3..5344eb8ca9 100644
--- a/libavfilter/Makefile
+++ b/libavfilter/Makefile
@@ -281,6 +281,7 @@  OBJS-$(CONFIG_HWUPLOAD_CUDA_FILTER)          += vf_hwupload_cuda.o
 OBJS-$(CONFIG_HWUPLOAD_FILTER)               += vf_hwupload.o
 OBJS-$(CONFIG_HYSTERESIS_FILTER)             += vf_hysteresis.o framesync.o
 OBJS-$(CONFIG_IDET_FILTER)                   += vf_idet.o
+OBJS-$(CONFIG_IIR_FILTER)                    += vf_iir.o
 OBJS-$(CONFIG_IL_FILTER)                     += vf_il.o
 OBJS-$(CONFIG_INFLATE_FILTER)                += vf_neighbor.o
 OBJS-$(CONFIG_INTERLACE_FILTER)              += vf_tinterlace.o
diff --git a/libavfilter/allfilters.c b/libavfilter/allfilters.c
index f2a44b0090..f6ecf96b9b 100644
--- a/libavfilter/allfilters.c
+++ b/libavfilter/allfilters.c
@@ -266,6 +266,7 @@  extern AVFilter ff_vf_hwupload;
 extern AVFilter ff_vf_hwupload_cuda;
 extern AVFilter ff_vf_hysteresis;
 extern AVFilter ff_vf_idet;
+extern AVFilter ff_vf_iir;
 extern AVFilter ff_vf_il;
 extern AVFilter ff_vf_inflate;
 extern AVFilter ff_vf_interlace;
diff --git a/libavfilter/vf_iir.c b/libavfilter/vf_iir.c
new file mode 100644
index 0000000000..193856adf3
--- /dev/null
+++ b/libavfilter/vf_iir.c
@@ -0,0 +1,410 @@ 
+/*
+ * Copyright (c) 2020 Paul B Mahol
+ *
+ * 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
+ */
+
+#include "libavutil/imgutils.h"
+#include "libavutil/opt.h"
+#include "libavutil/pixdesc.h"
+#include "libavutil/avstring.h"
+#include "avfilter.h"
+#include "formats.h"
+#include "internal.h"
+#include "video.h"
+
+typedef struct IIRContext {
+    const AVClass *class;
+
+    char *h_num_str;
+    char *h_den_str;
+    char *v_num_str;
+    char *v_den_str;
+
+    int h_nb_num;
+    int h_nb_den;
+    int v_nb_num;
+    int v_nb_den;
+
+    float *hnum;
+    float *hden;
+    float *vnum;
+    float *vden;
+
+    float **honum;
+    float **hoden;
+    float **vonum;
+    float **voden;
+
+    int planes;
+    int nb_planes;
+    int nb_threads;
+
+    int planewidth[4];
+    int planeheight[4];
+} IIRContext;
+
+#define OFFSET(x) offsetof(IIRContext, x)
+#define FLAGS AV_OPT_FLAG_VIDEO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
+
+static const AVOption iir_options[] = {
+    { "hn",     "set horizontal numerator",   OFFSET(h_num_str), AV_OPT_TYPE_STRING,   {.str=0},     0,   0, FLAGS },
+    { "hd",     "set horizotnal denominator", OFFSET(h_den_str), AV_OPT_TYPE_STRING,   {.str=0},     0,   0, FLAGS },
+    { "vn",     "set vertical numerator",     OFFSET(v_num_str), AV_OPT_TYPE_STRING,   {.str=0},     0,   0, FLAGS },
+    { "vd",     "set vertical denominator",   OFFSET(v_den_str), AV_OPT_TYPE_STRING,   {.str=0},     0,   0, FLAGS },
+    { "planes", "set planes to filter",       OFFSET(planes),    AV_OPT_TYPE_INT,      {.i64=0xF},   0, 0xF, FLAGS },
+    { NULL }
+};
+
+AVFILTER_DEFINE_CLASS(iir);
+
+typedef struct ThreadData {
+    int height;
+    int width;
+
+    float *dst;
+    float *src;
+    int dst_linesize;
+    int src_linesize;
+} ThreadData;
+
+static void horiz_slice(float *dst, float *src,
+                        int dst_linesize, int src_linesize,
+                        int nb_num, int nb_den,
+                        const float *num, const float *den,
+                        float *onum, float *oden,
+                        int width, int height)
+{
+    for (int y = 0; y < height; y++) {
+        for (int i = 0; i < nb_num; i++)
+            onum[i] = src[0];
+        for (int i = 0; i < nb_den; i++)
+            oden[i] = src[0];
+
+        for (int x = 0; x < width; x++) {
+            float sample = 0.f;
+
+            memmove(&onum[1], &onum[0], (nb_num - 1) * sizeof(*onum));
+            memmove(&oden[1], &oden[0], (nb_den - 1) * sizeof(*oden));
+            onum[0] = src[x];
+            for (int i = 0; i < nb_num; i++)
+                sample += num[i] * onum[i];
+
+            for (int i = 1; i < nb_den; i++)
+                sample -= den[i] * oden[i];
+
+            dst[x] = oden[0] = sample;
+        }
+
+        dst += dst_linesize;
+        src += src_linesize;
+    }
+}
+
+static int filter_horizontally(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
+{
+    IIRContext *s = ctx->priv;
+    ThreadData *td = arg;
+    const int height = td->height;
+    const int width = td->width;
+    const int slice_start = (height *  jobnr   ) / nb_jobs;
+    const int slice_end   = (height * (jobnr+1)) / nb_jobs;
+
+    horiz_slice(td->dst + td->dst_linesize * slice_start,
+                td->src + td->src_linesize * slice_start,
+                td->dst_linesize, td->src_linesize,
+                s->h_nb_num, s->h_nb_den,
+                s->hnum, s->hden, s->honum[jobnr], s->hoden[jobnr],
+                width, slice_end - slice_start);
+
+    return 0;
+}
+
+static void do_vertical_columns(float *pdst, float *psrc,
+                                int dst_linesize, int src_linesize,
+                                int nb_num, int nb_den,
+                                const float *num, const float *den,
+                                float *onum, float *oden,
+                                int width, int height)
+{
+    for (int x = 0; x < width; x++) {
+        float *src = psrc;
+        float *dst = pdst;
+
+        for (int i = 0; i < nb_num; i++)
+            onum[i] = src[x];
+        for (int i = 0; i < nb_den; i++)
+            oden[i] = src[x];
+
+        for (int y = 0; y < height; y++) {
+            float sample = 0.f;
+
+            memmove(&onum[1], &onum[0], (nb_num - 1) * sizeof(*onum));
+            memmove(&oden[1], &oden[0], (nb_den - 1) * sizeof(*oden));
+            onum[0] = src[x];
+            for (int i = 0; i < nb_num; i++)
+                sample += num[i] * onum[i];
+
+            for (int i = 1; i < nb_den; i++)
+                sample -= den[i] * oden[i];
+
+            dst[x] = oden[0] = sample;
+
+            dst += dst_linesize;
+            src += src_linesize;
+        }
+    }
+}
+
+static int filter_vertically(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
+{
+    IIRContext *s = ctx->priv;
+    ThreadData *td = arg;
+    const int height = td->height;
+    const int width = td->width;
+    const int slice_start = (width *  jobnr   ) / nb_jobs;
+    const int slice_end   = (width * (jobnr+1)) / nb_jobs;
+
+    do_vertical_columns(td->dst + slice_start,
+                        td->src + slice_start,
+                        td->dst_linesize, td->src_linesize,
+                        s->v_nb_num, s->v_nb_den,
+                        s->vnum, s->vden, s->vonum[jobnr], s->voden[jobnr],
+                        slice_end - slice_start, height);
+
+    return 0;
+}
+
+static void iir2d(AVFilterContext *ctx, AVFrame *out, AVFrame *in, int plane)
+{
+    IIRContext *s = ctx->priv;
+    const int width = s->planewidth[plane];
+    const int height = s->planeheight[plane];
+    ThreadData td;
+
+    td.width = width;
+    td.height = height;
+    td.dst = (float *)out->data[plane];
+    td.src = (float *)in->data[plane];
+    td.dst_linesize = out->linesize[plane] / 4;
+    td.src_linesize = in->linesize[plane] / 4;
+    ctx->internal->execute(ctx, filter_horizontally, &td, NULL, FFMIN(height, s->nb_threads));
+    td.src = (float *)out->data[plane];
+    td.src_linesize = out->linesize[plane] / 4;
+    ctx->internal->execute(ctx, filter_vertically, &td, NULL, FFMIN(width, s->nb_threads));
+}
+
+static int query_formats(AVFilterContext *ctx)
+{
+    static const enum AVPixelFormat pix_fmts[] = {
+        AV_PIX_FMT_GBRPF32, AV_PIX_FMT_GBRAPF32,
+        AV_PIX_FMT_NONE
+    };
+
+    return ff_set_common_formats(ctx, ff_make_format_list(pix_fmts));
+}
+
+static void count_coefficients(char *item_str, int *nb_items)
+{
+    char *p;
+
+    if (!item_str)
+        return;
+
+    *nb_items = 1;
+    for (p = item_str; *p; p++) {
+        if (*p == ' ')
+            (*nb_items)++;
+    }
+}
+
+static int read_coefficients(AVFilterContext *ctx, char *item_str, int nb_items, float *dst)
+{
+    char *p, *arg, *old_str, *saveptr = NULL;
+    int i;
+
+    p = old_str = av_strdup(item_str);
+    if (!p)
+        return AVERROR(ENOMEM);
+
+    for (i = 0; i < nb_items; i++) {
+        if (!(arg = av_strtok(p, " ", &saveptr)))
+            break;
+
+        p = NULL;
+        if (av_sscanf(arg, "%f", &dst[i]) != 1) {
+            av_log(ctx, AV_LOG_ERROR, "Invalid coefficients supplied: %s\n", arg);
+            av_freep(&old_str);
+            return AVERROR(EINVAL);
+        }
+    }
+
+    av_freep(&old_str);
+
+    return 0;
+}
+
+static int config_input(AVFilterLink *inlink)
+{
+    const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(inlink->format);
+    AVFilterContext *ctx = inlink->dst;
+    IIRContext *s = ctx->priv;
+    int ret;
+
+    s->nb_threads = ff_filter_get_nb_threads(ctx);
+    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);
+
+    count_coefficients(s->h_num_str, &s->h_nb_num);
+    count_coefficients(s->h_den_str, &s->h_nb_den);
+    count_coefficients(s->v_num_str, &s->v_nb_num);
+    count_coefficients(s->v_den_str, &s->v_nb_den);
+
+    if (!s->h_nb_num || !s->h_nb_den ||
+        !s->v_nb_num || !s->v_nb_den)
+        return AVERROR(EINVAL);
+
+    s->hnum  = av_calloc(s->h_nb_num, sizeof(*s->hnum));
+    s->hden  = av_calloc(s->h_nb_den, sizeof(*s->hden));
+    s->honum = av_calloc(s->nb_threads, sizeof(*s->honum));
+    s->hoden = av_calloc(s->nb_threads, sizeof(*s->hoden));
+    s->vnum  = av_calloc(s->h_nb_num, sizeof(*s->vnum));
+    s->vden  = av_calloc(s->h_nb_den, sizeof(*s->vden));
+    s->vonum = av_calloc(s->nb_threads, sizeof(*s->vonum));
+    s->voden = av_calloc(s->nb_threads, sizeof(*s->voden));
+    if (!s->hnum || !s->hden || !s->honum || !s->hoden ||
+        !s->vnum || !s->vden || !s->vonum || !s->voden)
+        return AVERROR(ENOMEM);
+
+    for (int i = 0; i < s->nb_threads; i++) {
+        s->honum[i] = av_calloc(s->h_nb_num, sizeof(**s->honum));
+        s->hoden[i] = av_calloc(s->h_nb_den, sizeof(**s->hoden));
+        s->vonum[i] = av_calloc(s->v_nb_num, sizeof(**s->vonum));
+        s->voden[i] = av_calloc(s->v_nb_den, sizeof(**s->voden));
+        if (!s->honum[i] || !s->hoden[i] ||
+            !s->vonum[i] || !s->voden[i])
+            return AVERROR(ENOMEM);
+    }
+
+    ret = read_coefficients(ctx, s->h_num_str, s->h_nb_num, s->hnum);
+    if (ret < 0)
+        return ret;
+    ret = read_coefficients(ctx, s->h_den_str, s->h_nb_den, s->hden);
+    if (ret < 0)
+        return ret;
+    ret = read_coefficients(ctx, s->v_num_str, s->v_nb_num, s->vnum);
+    if (ret < 0)
+        return ret;
+    ret = read_coefficients(ctx, s->v_den_str, s->v_nb_den, s->vden);
+    if (ret < 0)
+        return ret;
+
+    return 0;
+}
+
+static int filter_frame(AVFilterLink *inlink, AVFrame *in)
+{
+    AVFilterContext *ctx = inlink->dst;
+    IIRContext *s = ctx->priv;
+    AVFilterLink *outlink = ctx->outputs[0];
+    AVFrame *out;
+    int plane;
+
+    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++) {
+        const int height = s->planeheight[plane];
+        const int width = s->planewidth[plane];
+
+        if (!(s->planes & (1 << plane))) {
+            av_image_copy_plane(out->data[plane], out->linesize[plane],
+                                in->data[plane], in->linesize[plane],
+                                width * sizeof(float), height);
+            continue;
+        }
+
+        iir2d(ctx, out, in, plane);
+    }
+
+    av_frame_free(&in);
+    return ff_filter_frame(outlink, out);
+}
+
+static av_cold void uninit(AVFilterContext *ctx)
+{
+    IIRContext *s = ctx->priv;
+
+    av_freep(&s->hnum);
+    av_freep(&s->hden);
+    av_freep(&s->vnum);
+    av_freep(&s->vden);
+
+    for (int i = 0; i < s->nb_threads; i++) {
+        if (s->honum)
+            av_freep(&s->honum[i]);
+        if (s->hoden)
+            av_freep(&s->hoden[i]);
+        if (s->vonum)
+            av_freep(&s->vonum[i]);
+        if (s->voden)
+            av_freep(&s->voden[i]);
+    }
+
+    av_freep(&s->honum);
+    av_freep(&s->hoden);
+    av_freep(&s->vonum);
+    av_freep(&s->voden);
+}
+
+static const AVFilterPad iir_inputs[] = {
+    {
+        .name         = "default",
+        .type         = AVMEDIA_TYPE_VIDEO,
+        .config_props = config_input,
+        .filter_frame = filter_frame,
+    },
+    { NULL }
+};
+
+static const AVFilterPad iir_outputs[] = {
+    {
+        .name = "default",
+        .type = AVMEDIA_TYPE_VIDEO,
+    },
+    { NULL }
+};
+
+AVFilter ff_vf_iir = {
+    .name          = "iir",
+    .description   = NULL_IF_CONFIG_SMALL("Apply 2D Infinite Impulse Response filter with supplied coefficients."),
+    .priv_size     = sizeof(IIRContext),
+    .priv_class    = &iir_class,
+    .uninit        = uninit,
+    .query_formats = query_formats,
+    .inputs        = iir_inputs,
+    .outputs       = iir_outputs,
+    .flags         = AVFILTER_FLAG_SUPPORT_TIMELINE_GENERIC | AVFILTER_FLAG_SLICE_THREADS,
+};