avfilter: add median filter

8 files changed, 503 insertions, 1 deletions

diff --git a/Changelog b/Changelog
index 6fab938e0f..7c971fce63 100644
--- a/Changelog
+++ b/Changelog
@@ -19,6 +19,7 @@ version <next>:
 - bilateral filter
 - maskedmin and maskedmax filters
 - VDPAU VP9 hwaccel
+- median filter
 
 
 version 4.2:
diff --git a/doc/filters.texi b/doc/filters.texi
index 5b8062d806..c4477e6677 100644
--- a/doc/filters.texi
+++ b/doc/filters.texi
@@ -12805,6 +12805,21 @@ Higher values should result in a smoother motion vector field but less
 optimal individual vectors. Default value is 1.
 @end table
 
+@section median
+
+Pick median pixel from certain rectangle defined by radius.
+
+This filter accepts the following options:
+
+@table @option
+@item radius
+Set horizontal and vertical radius size. Default value is @code{1}.
+Allowed range is integer from 1 to 127.
+
+@item planes
+Set which planes to process. Default is @code{15}, which is all available planes.
+@end table
+
 @section mergeplanes
 
 Merge color channel components from several video streams.
diff --git a/libavfilter/Makefile b/libavfilter/Makefile
index bae566d433..2080eed559 100644
--- a/libavfilter/Makefile
+++ b/libavfilter/Makefile
@@ -297,6 +297,7 @@ OBJS-$(CONFIG_MASKEDMERGE_FILTER)            += vf_maskedmerge.o framesync.o
 OBJS-$(CONFIG_MASKEDMIN_FILTER)              += vf_maskedminmax.o framesync.o
 OBJS-$(CONFIG_MASKFUN_FILTER)                += vf_maskfun.o
 OBJS-$(CONFIG_MCDEINT_FILTER)                += vf_mcdeint.o
+OBJS-$(CONFIG_MEDIAN_FILTER)                 += vf_median.o
 OBJS-$(CONFIG_MERGEPLANES_FILTER)            += vf_mergeplanes.o framesync.o
 OBJS-$(CONFIG_MESTIMATE_FILTER)              += vf_mestimate.o motion_estimation.o
 OBJS-$(CONFIG_METADATA_FILTER)               += f_metadata.o
diff --git a/libavfilter/allfilters.c b/libavfilter/allfilters.c
index d136734338..cb609067b6 100644
--- a/libavfilter/allfilters.c
+++ b/libavfilter/allfilters.c
@@ -282,6 +282,7 @@ extern AVFilter ff_vf_maskedmerge;
 extern AVFilter ff_vf_maskedmin;
 extern AVFilter ff_vf_maskfun;
 extern AVFilter ff_vf_mcdeint;
+extern AVFilter ff_vf_median;
 extern AVFilter ff_vf_mergeplanes;
 extern AVFilter ff_vf_mestimate;
 extern AVFilter ff_vf_metadata;
diff --git a/libavfilter/median.h b/libavfilter/median.h
new file mode 100644
index 0000000000..bb07fddd5d
--- /dev/null
+++ b/libavfilter/median.h
@@ -0,0 +1,54 @@
+/*
+ * Copyright (c) 2019 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
+ * Redistribution and use in source and binary forms, with or without modification,
+ * are permitted provided that the following conditions are met:
+ */
+
+#ifndef AVFILTER_MEDIAN_H
+#define AVFILTER_MEDIAN_H
+
+#include "avfilter.h"
+
+typedef struct MedianContext {
+    const AVClass *class;
+
+    int planes;
+    int radius;
+
+    int planewidth[4];
+    int planeheight[4];
+    int depth;
+    int nb_planes;
+    int nb_threads;
+
+    uint16_t **coarse, **fine;
+    int coarse_size, fine_size;
+    int bins;
+    int t;
+
+    void (*hadd)(uint16_t *dst, const uint16_t *src, int bins);
+    void (*hsub)(uint16_t *dst, const uint16_t *src, int bins);
+    void (*hmuladd)(uint16_t *dst, const uint16_t *src, int f, int bins);
+
+    void (*filter_plane)(AVFilterContext *ctx, const uint8_t *ssrc, int src_linesize,
+                         uint8_t *ddst, int dst_linesize, int width, int height,
+                         int slice_h_start, int slice_h_end, int jobnr);
+} MedianContext;
+
+#endif
diff --git a/libavfilter/median_template.c b/libavfilter/median_template.c
new file mode 100644
index 0000000000..6953a23a5b
--- /dev/null
+++ b/libavfilter/median_template.c
@@ -0,0 +1,159 @@
+/*
+ * Copyright (c) 2019 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
+ * Redistribution and use in source and binary forms, with or without modification,
+ * are permitted provided that the following conditions are met:
+ */
+
+#include "libavutil/avassert.h"
+#include "avfilter.h"
+#include "formats.h"
+#include "internal.h"
+#include "video.h"
+
+#undef pixel
+#if DEPTH == 8
+#define pixel uint8_t
+#else
+#define pixel uint16_t
+#endif
+
+#undef htype
+#define htype uint16_t
+
+#undef fn
+#undef fn2
+#undef fn3
+#define SHIFT   ((DEPTH + 1) / 2)
+#define BINS    (1 << SHIFT)
+#define MASK    (BINS - 1)
+#define fn3(a,b)   a##_##b
+#define fn2(a,b)   fn3(a,b)
+#define fn(a)      fn2(a, DEPTH)
+
+#define PICK_COARSE_BIN(x, y) (BINS * (x) + ((y) >> SHIFT))
+#define PICK_FINE_BIN(x, y, z) (BINS * ((x) * ((y) >> SHIFT) + (z)) + ((y) & MASK))
+
+static void fn(filter_plane)(AVFilterContext *ctx, const uint8_t *ssrc, int src_linesize,
+                             uint8_t *ddst, int dst_linesize, int width, int height,
+                             int slice_h_start, int slice_h_end, int jobnr)
+{
+    MedianContext *s = ctx->priv;
+    htype *ccoarse = s->coarse[jobnr];
+    htype *cfine = s->fine[jobnr];
+    const int radius = s->radius;
+    const int t = s->t;
+    const pixel *src = (const pixel *)ssrc;
+    pixel *dst = (pixel *)ddst;
+    const pixel *srcp;
+    const pixel *p;
+
+    src_linesize /= sizeof(pixel);
+    dst_linesize /= sizeof(pixel);
+
+    memset(cfine, 0, s->fine_size * sizeof(*cfine));
+    memset(ccoarse, 0, s->coarse_size * sizeof(*ccoarse));
+
+    srcp = src + FFMAX(0, slice_h_start - radius) * src_linesize;
+    if (jobnr == 0) {
+        for (int i = 0; i < width; i++) {
+            cfine[PICK_FINE_BIN(width, srcp[i], i)] += radius + 1;
+            ccoarse[PICK_COARSE_BIN(i, srcp[i])] += radius + 1;
+        }
+    }
+
+    srcp = src + FFMAX(0, slice_h_start - radius - (jobnr != 0)) * src_linesize;
+    for (int i = 0; i < radius + (jobnr != 0) * (1 + radius); i++) {
+        for (int j = 0; j < width; j++) {
+            cfine[PICK_FINE_BIN(width, srcp[j], j)]++;
+            ccoarse[PICK_COARSE_BIN(j, srcp[j])]++;
+        }
+        srcp += src_linesize;
+    }
+
+    srcp = src;
+
+    for (int i = slice_h_start; i < slice_h_end; i++) {
+        htype coarse[BINS] = { 0 };
+        htype fine[BINS][BINS] = { { 0 } };
+        htype luc[BINS] = { 0 };
+
+        p = srcp + src_linesize * FFMAX(0, i - radius - 1);
+        for (int j = 0; j < width; j++) {
+            cfine[PICK_FINE_BIN(width, p[j], j)]--;
+            ccoarse[PICK_COARSE_BIN(j, p[j])]--;
+        }
+
+        p = srcp + src_linesize * FFMIN(height - 1, i + radius);
+        for (int j = 0; j < width; j++) {
+            cfine[PICK_FINE_BIN(width, p[j], j)]++;
+            ccoarse[PICK_COARSE_BIN(j, p[j])]++;
+        }
+
+        s->hmuladd(coarse, &ccoarse[0], radius, BINS);
+        for (int j = 0; j < radius; j++)
+            s->hadd(coarse, &ccoarse[BINS * j], BINS);
+        for (int k = 0; k < BINS; k++)
+            s->hmuladd(&fine[k][0], &cfine[BINS * width * k], 2 * radius + 1, BINS);
+
+        for (int j = 0; j < width; j++) {
+            int sum = 0, k, b;
+            htype *segment;
+
+            s->hadd(coarse, &ccoarse[BINS * FFMIN(j + radius, width - 1)], BINS);
+
+            for (k = 0; k < BINS; k++) {
+                sum += coarse[k];
+                if (sum > t) {
+                    sum -= coarse[k];
+                    break;
+                }
+            }
+            av_assert0(k < BINS);
+
+            if (luc[k] <= j - radius) {
+                memset(&fine[k], 0, BINS * sizeof(htype));
+                for (luc[k] = j - radius; luc[k] < FFMIN(j + radius + 1, width); luc[k]++)
+                    s->hadd(fine[k], &cfine[BINS * (width * k + luc[k])], BINS);
+                if (luc[k] < j + radius + 1) {
+                    s->hmuladd(&fine[k][0], &cfine[BINS * (width * k + width - 1)], j + radius + 1 - width, BINS);
+                    luc[k] = j + radius + 1;
+                }
+            } else {
+                for (; luc[k] < j + radius + 1; luc[k]++) {
+                    s->hsub(fine[k], &cfine[BINS * (width * k + FFMAX(luc[k] - 2 * radius - 1, 0))], BINS);
+                    s->hadd(fine[k], &cfine[BINS * (width * k + FFMIN(luc[k], width - 1))], BINS);
+                }
+            }
+
+            s->hsub(coarse, &ccoarse[BINS * FFMAX(j - radius, 0)], BINS);
+
+            segment = fine[k];
+            for (b = 0; b < BINS; b++) {
+                sum += segment[b];
+                if (sum > t) {
+                    dst[j] = BINS * k + b;
+                    break;
+                }
+            }
+            av_assert0(b < BINS);
+        }
+
+        dst += dst_linesize;
+    }
+}
diff --git a/libavfilter/version.h b/libavfilter/version.h
index 58d57ae95f..2e3ff53b20 100644
--- a/libavfilter/version.h
+++ b/libavfilter/version.h
@@ -30,7 +30,7 @@
 #include "libavutil/version.h"
 
 #define LIBAVFILTER_VERSION_MAJOR   7
-#define LIBAVFILTER_VERSION_MINOR  65
+#define LIBAVFILTER_VERSION_MINOR  66
 #define LIBAVFILTER_VERSION_MICRO 100
 
 
diff --git a/libavfilter/vf_median.c b/libavfilter/vf_median.c
new file mode 100644
index 0000000000..98296bcc2b
--- /dev/null
+++ b/libavfilter/vf_median.c
@@ -0,0 +1,271 @@
+/*
+ * Copyright (c) 2019 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
+ * Redistribution and use in source and binary forms, with or without modification,
+ * are permitted provided that the following conditions are met:
+ */
+
+#include "libavutil/avassert.h"
+#include "libavutil/imgutils.h"
+#include "libavutil/opt.h"
+#include "libavutil/pixdesc.h"
+#include "avfilter.h"
+#include "formats.h"
+#include "internal.h"
+#include "median.h"
+#include "video.h"
+
+#define DEPTH 8
+#include "median_template.c"
+
+#undef DEPTH
+#define DEPTH 9
+#include "median_template.c"
+
+#undef DEPTH
+#define DEPTH 10
+#include "median_template.c"
+
+#undef DEPTH
+#define DEPTH 12
+#include "median_template.c"
+
+#undef DEPTH
+#define DEPTH 14
+#include "median_template.c"
+
+#undef DEPTH
+#define DEPTH 16
+#include "median_template.c"
+
+#define OFFSET(x) offsetof(MedianContext, x)
+#define FLAGS AV_OPT_FLAG_VIDEO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
+
+static const AVOption median_options[] = {
+    { "radius", "set median radius",    OFFSET(radius), AV_OPT_TYPE_INT,   {.i64=1},     1,  127, FLAGS },
+    { "planes", "set planes to filter", OFFSET(planes), AV_OPT_TYPE_INT,   {.i64=0xF},   0,  0xF, FLAGS },
+    { NULL }
+};
+
+AVFILTER_DEFINE_CLASS(median);
+
+static void hadd(htype *dst, const htype *src, int bins)
+{
+    for (int i = 0; i < bins; i++)
+        dst[i] += src[i];
+}
+
+static void hsub(htype *dst, const htype *src, int bins)
+{
+    for (int i = 0; i < bins; i++)
+        dst[i] -= src[i];
+}
+
+static void hmuladd(htype *dst, const htype *src, int f, int bins)
+{
+    for (int i = 0; i < bins; i++)
+        dst[i] += f * src[i];
+}
+
+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_GBRP, AV_PIX_FMT_GBRAP, AV_PIX_FMT_GRAY8, AV_PIX_FMT_GRAY9,
+        AV_PIX_FMT_YUV420P9, AV_PIX_FMT_YUV422P9, AV_PIX_FMT_YUV444P9, AV_PIX_FMT_GBRP9,
+        AV_PIX_FMT_YUVA420P9, AV_PIX_FMT_YUVA422P9, AV_PIX_FMT_YUVA444P9,
+        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_YUVA420P10, AV_PIX_FMT_YUVA422P10, AV_PIX_FMT_YUVA444P10,
+        AV_PIX_FMT_YUVA420P16, AV_PIX_FMT_YUVA422P16, AV_PIX_FMT_YUVA444P16,
+        AV_PIX_FMT_GBRP10, AV_PIX_FMT_GBRP12, AV_PIX_FMT_GBRP14, AV_PIX_FMT_GBRP16,
+        AV_PIX_FMT_GBRAP10, AV_PIX_FMT_GBRAP12, AV_PIX_FMT_GBRAP16,
+        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));
+}
+
+static int config_input(AVFilterLink *inlink)
+{
+    const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(inlink->format);
+    MedianContext *s = inlink->dst->priv;
+
+    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->t = 2 * s->radius * s->radius + 2 * s->radius;
+
+    for (int i = 0; i < s->nb_planes; i++) {
+        if (!(s->planes & (1 << i)))
+            continue;
+
+        if (s->planewidth[i] < s->radius * 2 + 1) {
+            av_log(inlink->dst, AV_LOG_ERROR, "The %d plane width %d must be not less than %d\n", i, s->planewidth[i], s->radius * 2 + 1);
+            return AVERROR(EINVAL);
+        }
+
+        if (s->planeheight[i] < s->radius * 2 + 1) {
+            av_log(inlink->dst, AV_LOG_ERROR, "The %d plane height %d must be not less than %d\n", i, s->planeheight[i], s->radius * 2 + 1);
+            return AVERROR(EINVAL);
+        }
+    }
+
+    s->nb_threads = FFMAX(1, FFMIN(s->planeheight[1] / (s->radius + 1), ff_filter_get_nb_threads(inlink->dst)));
+    s->bins   = 1 << ((s->depth + 1) / 2);
+    s->fine_size = s->bins * s->bins * inlink->w;
+    s->coarse_size = s->bins * inlink->w;
+    s->coarse = av_calloc(s->nb_threads, sizeof(*s->coarse));
+    s->fine   = av_calloc(s->nb_threads, sizeof(*s->fine));
+    if (!s->coarse || !s->fine)
+        return AVERROR(ENOMEM);
+    for (int i = 0; i < s->nb_threads; i++) {
+        s->coarse[i] = av_malloc_array(s->coarse_size, sizeof(**s->coarse));
+        s->fine[i]   = av_malloc_array(s->fine_size, sizeof(**s->fine));
+        if (!s->coarse[i] || !s->fine[i])
+            return AVERROR(ENOMEM);
+    }
+
+    s->hadd = hadd;
+    s->hsub = hsub;
+    s->hmuladd = hmuladd;
+
+    switch (s->depth) {
+    case  8: s->filter_plane = filter_plane_8;  break;
+    case  9: s->filter_plane = filter_plane_9;  break;
+    case 10: s->filter_plane = filter_plane_10; break;
+    case 12: s->filter_plane = filter_plane_12; break;
+    case 14: s->filter_plane = filter_plane_14; break;
+    case 16: s->filter_plane = filter_plane_16; break;
+    }
+
+    return 0;
+}
+
+typedef struct ThreadData {
+    AVFrame *in, *out;
+} ThreadData;
+
+static int filter_slice(AVFilterContext *ctx, void *arg, int jobnr, int nb_jobs)
+{
+    MedianContext *s = ctx->priv;
+    ThreadData *td = arg;
+    AVFrame *in = td->in;
+    AVFrame *out = td->out;
+
+    for (int plane = 0; plane < s->nb_planes; plane++) {
+        const int h = s->planeheight[plane];
+        const int w = s->planewidth[plane];
+        const int slice_h_start = (h * jobnr) / nb_jobs;
+        const int slice_h_end = (h * (jobnr+1)) / nb_jobs;
+
+        if (!(s->planes & (1 << plane))) {
+            av_image_copy_plane(out->data[plane] + slice_h_start * out->linesize[plane],
+                                out->linesize[plane],
+                                in->data[plane] + slice_h_start * in->linesize[plane],
+                                in->linesize[plane],
+                                w * ((s->depth + 7) / 8),
+                                slice_h_end - slice_h_start);
+            continue;
+        }
+
+        s->filter_plane(ctx, in->data[plane],
+                        in->linesize[plane],
+                        out->data[plane] + slice_h_start * out->linesize[plane],
+                        out->linesize[plane], w, h,
+                        slice_h_start, slice_h_end, jobnr);
+    }
+
+    return 0;
+}
+
+static int filter_frame(AVFilterLink *inlink, AVFrame *in)
+{
+    AVFilterContext *ctx = inlink->dst;
+    MedianContext *s = ctx->priv;
+    AVFilterLink *outlink = ctx->outputs[0];
+    ThreadData td;
+    AVFrame *out;
+
+    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);
+
+    td.in = in; td.out = out;
+    ctx->internal->execute(ctx, filter_slice, &td, NULL, s->nb_threads);
+
+    av_frame_free(&in);
+    return ff_filter_frame(outlink, out);
+}
+
+static av_cold void uninit(AVFilterContext *ctx)
+{
+    MedianContext *s = ctx->priv;
+
+    for (int i = 0; i < s->nb_threads && s->coarse && s->fine; i++) {
+        av_freep(&s->coarse[i]);
+        av_freep(&s->fine[i]);
+    }
+
+    av_freep(&s->coarse);
+    av_freep(&s->fine);
+}
+
+static const AVFilterPad median_inputs[] = {
+    {
+        .name         = "default",
+        .type         = AVMEDIA_TYPE_VIDEO,
+        .config_props = config_input,
+        .filter_frame = filter_frame,
+    },
+    { NULL }
+};
+
+static const AVFilterPad median_outputs[] = {
+    {
+        .name = "default",
+        .type = AVMEDIA_TYPE_VIDEO,
+    },
+    { NULL }
+};
+
+AVFilter ff_vf_median = {
+    .name          = "median",
+    .description   = NULL_IF_CONFIG_SMALL("Apply Median filter."),
+    .priv_size     = sizeof(MedianContext),
+    .priv_class    = &median_class,
+    .uninit        = uninit,
+    .query_formats = query_formats,
+    .inputs        = median_inputs,
+    .outputs       = median_outputs,
+    .flags         = AVFILTER_FLAG_SUPPORT_TIMELINE_GENERIC | AVFILTER_FLAG_SLICE_THREADS,
+};