[FFmpeg-devel] [PATCH] avfilter: add estdif video filter

Paul B Mahol onemda at gmail.com
Wed Jan 13 16:04:31 EET 2021


Signed-off-by: Paul B Mahol <onemda at gmail.com>
---
 doc/filters.texi         |  64 ++++++
 libavfilter/Makefile     |   1 +
 libavfilter/allfilters.c |   1 +
 libavfilter/vf_estdif.c  | 469 +++++++++++++++++++++++++++++++++++++++
 4 files changed, 535 insertions(+)
 create mode 100644 libavfilter/vf_estdif.c

diff --git a/doc/filters.texi b/doc/filters.texi
index 813e35c2f9..9fbc4de7d2 100644
--- a/doc/filters.texi
+++ b/doc/filters.texi
@@ -11065,6 +11065,70 @@ Flags to local 3x3 coordinates maps like this:
 
 This filter supports the all above options as @ref{commands}.
 
+ at section estdif
+
+Deinterlace the input video ("estdif" stands for "Edge Slope
+Tracing Deinterlacing Filter").
+
+Spatial only filter that uses edge slope tracing algorithm
+to interpolate missing lines.
+It accepts the following parameters:
+
+ at table @option
+ at item mode
+The interlacing mode to adopt. It accepts one of the following values:
+
+ at table @option
+ at item frame
+Output one frame for each frame.
+ at item field
+Output one frame for each field.
+ at end table
+
+The default value is @code{field}.
+
+ at item parity
+The picture field parity assumed for the input interlaced video. It accepts one
+of the following values:
+
+ at table @option
+ at item ff
+Assume the top field is first.
+ at item bff
+Assume the bottom field is first.
+ at item auto
+Enable automatic detection of field parity.
+ at end table
+
+The default value is @code{auto}.
+If the interlacing is unknown or the decoder does not export this information,
+top field first will be assumed.
+
+ at item deint
+Specify which frames to deinterlace. Accepts one of the following
+values:
+
+ at table @option
+ at item all
+Deinterlace all frames.
+ at item interlaced
+Only deinterlace frames marked as interlaced.
+ at end table
+
+The default value is @code{all}.
+
+ at item rslope
+Specify the search radius for edge slope tracing. Default value is 1.
+Allowed range is from 1 to 15.
+
+ at item redge
+Specify the search radius for best edge matching. Default value is 2.
+Allowed range is from 0 to 3.
+ at end table
+
+ at subsection Commands
+This filter supports same @ref{commands} as options.
+
 @section extractplanes
 
 Extract color channel components from input video stream into
diff --git a/libavfilter/Makefile b/libavfilter/Makefile
index ad1046d526..44afa79963 100644
--- a/libavfilter/Makefile
+++ b/libavfilter/Makefile
@@ -252,6 +252,7 @@ OBJS-$(CONFIG_EQ_FILTER)                     += vf_eq.o
 OBJS-$(CONFIG_EROSION_FILTER)                += vf_neighbor.o
 OBJS-$(CONFIG_EROSION_OPENCL_FILTER)         += vf_neighbor_opencl.o opencl.o \
                                                 opencl/neighbor.o
+OBJS-$(CONFIG_ESTDIF_FILTER)                 += vf_estdif.o
 OBJS-$(CONFIG_EXTRACTPLANES_FILTER)          += vf_extractplanes.o
 OBJS-$(CONFIG_FADE_FILTER)                   += vf_fade.o
 OBJS-$(CONFIG_FFTDNOIZ_FILTER)               += vf_fftdnoiz.o
diff --git a/libavfilter/allfilters.c b/libavfilter/allfilters.c
index ce317dfa1c..471844a603 100644
--- a/libavfilter/allfilters.c
+++ b/libavfilter/allfilters.c
@@ -237,6 +237,7 @@ extern AVFilter ff_vf_entropy;
 extern AVFilter ff_vf_eq;
 extern AVFilter ff_vf_erosion;
 extern AVFilter ff_vf_erosion_opencl;
+extern AVFilter ff_vf_estdif;
 extern AVFilter ff_vf_extractplanes;
 extern AVFilter ff_vf_fade;
 extern AVFilter ff_vf_fftdnoiz;
diff --git a/libavfilter/vf_estdif.c b/libavfilter/vf_estdif.c
new file mode 100644
index 0000000000..5d83e66906
--- /dev/null
+++ b/libavfilter/vf_estdif.c
@@ -0,0 +1,469 @@
+/*
+ * 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/common.h"
+#include "libavutil/imgutils.h"
+#include "libavutil/opt.h"
+#include "libavutil/pixdesc.h"
+#include "avfilter.h"
+#include "formats.h"
+#include "internal.h"
+#include "video.h"
+
+typedef struct ESTDIFContext {
+    const AVClass *class;
+
+    int mode;             ///< 0 is frame, 1 is field
+    int parity;           ///< frame field parity
+    int deint;            ///< which frames to deinterlace
+    int rslope;           ///< best edge slope search radius
+    int redge;            ///< best edge match search radius
+    int linesize[4];      ///< bytes of pixel data per line for each plane
+    int planewidth[4];    ///< width of each plane
+    int planeheight[4];   ///< height of each plane
+    int field;            ///< which field are we on, 0 or 1
+    int eof;
+    int nb_planes;
+    int nb_threads;
+    int64_t pts;
+    AVFrame *prev;
+
+    void (*interpolate)(uint8_t *dst,
+                        const uint8_t *prev_line, const uint8_t *next_line,
+                        const uint8_t *prev2_line, const uint8_t *next2_line,
+                        int x, int width, int rslope, int redge,
+                        int *K);
+} ESTDIFContext;
+
+#define MAX_R 15
+#define S     (2 * MAX_R + 1)
+
+#define OFFSET(x) offsetof(ESTDIFContext, x)
+#define FLAGS AV_OPT_FLAG_VIDEO_PARAM|AV_OPT_FLAG_FILTERING_PARAM|AV_OPT_FLAG_RUNTIME_PARAM
+#define CONST(name, help, val, unit) { name, help, 0, AV_OPT_TYPE_CONST, {.i64=val}, 0, 0, FLAGS, unit }
+
+static const AVOption estdif_options[] = {
+    { "mode", "specify the mode", OFFSET(mode), AV_OPT_TYPE_INT, {.i64=1}, 0, 1, FLAGS, "mode" },
+    CONST("frame", NULL, 0, "mode"),
+    CONST("field", NULL, 1, "mode"),
+    { "parity", "specify the assumed picture field parity", OFFSET(parity), AV_OPT_TYPE_INT, {.i64=-1}, -1, 1, FLAGS, "parity" },
+    CONST("tff",  "assume top field first",    0, "parity"),
+    CONST("bff",  "assume bottom field first", 1, "parity"),
+    CONST("auto", "auto detect parity",       -1, "parity"),
+    { "deint",  "specify which frames to deinterlace", OFFSET(deint), AV_OPT_TYPE_INT, {.i64=0}, 0, 1, FLAGS, "deint" },
+    CONST("all",        "deinterlace all frames",                       0, "deint"),
+    CONST("interlaced", "only deinterlace frames marked as interlaced", 1, "deint"),
+    { "rslope", "specify the search radius for edge slope tracing", OFFSET(rslope), AV_OPT_TYPE_INT, {.i64=1}, 1, MAX_R, FLAGS, },
+    { "redge",  "specify the search radius for best edge matching", OFFSET(redge),  AV_OPT_TYPE_INT, {.i64=2}, 0,     3, FLAGS, },
+    { NULL }
+};
+
+AVFILTER_DEFINE_CLASS(estdif);
+
+static int query_formats(AVFilterContext *ctx)
+{
+    static const enum AVPixelFormat pix_fmts[] = {
+        AV_PIX_FMT_YUV410P, AV_PIX_FMT_YUV411P,
+        AV_PIX_FMT_YUV420P, AV_PIX_FMT_YUV422P,
+        AV_PIX_FMT_YUV440P, AV_PIX_FMT_YUV444P,
+        AV_PIX_FMT_YUVJ444P, AV_PIX_FMT_YUVJ440P,
+        AV_PIX_FMT_YUVJ422P, AV_PIX_FMT_YUVJ420P,
+        AV_PIX_FMT_YUVJ411P,
+        AV_PIX_FMT_YUVA420P, AV_PIX_FMT_YUVA422P, AV_PIX_FMT_YUVA444P,
+        AV_PIX_FMT_GBRP, AV_PIX_FMT_GBRAP,
+        AV_PIX_FMT_GRAY8,
+        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_YUV420P14, AV_PIX_FMT_YUV422P14, AV_PIX_FMT_YUV444P14,
+        AV_PIX_FMT_GBRP9, AV_PIX_FMT_GBRP10, AV_PIX_FMT_GBRP12, AV_PIX_FMT_GBRP14,
+        AV_PIX_FMT_YUVA444P9, AV_PIX_FMT_YUVA444P10, AV_PIX_FMT_YUVA444P12, AV_PIX_FMT_YUVA444P16,
+        AV_PIX_FMT_YUVA422P9, AV_PIX_FMT_YUVA422P10, AV_PIX_FMT_YUVA422P12, AV_PIX_FMT_YUVA422P16,
+        AV_PIX_FMT_YUVA420P9, AV_PIX_FMT_YUVA420P10, AV_PIX_FMT_YUVA420P16,
+        AV_PIX_FMT_GBRAP10,   AV_PIX_FMT_GBRAP12,    AV_PIX_FMT_GBRAP16,
+        AV_PIX_FMT_NONE
+    };
+
+    AVFilterFormats *fmts_list = ff_make_format_list(pix_fmts);
+    if (!fmts_list)
+        return AVERROR(ENOMEM);
+    return ff_set_common_formats(ctx, fmts_list);
+}
+
+static int config_output(AVFilterLink *outlink)
+{
+    AVFilterContext *ctx = outlink->src;
+    AVFilterLink *inlink = ctx->inputs[0];
+
+    outlink->time_base.num = inlink->time_base.num;
+    outlink->time_base.den = inlink->time_base.den * 2;
+    outlink->frame_rate.num = inlink->frame_rate.num * 2;
+    outlink->frame_rate.den = inlink->frame_rate.den;
+
+    return 0;
+}
+
+typedef struct ThreadData {
+    AVFrame *out, *in;
+} ThreadData;
+
+#define MIDL(type, ss)                                    \
+static int midl_##ss(const type *prev, const type *next,  \
+                int end, int x, int k)                    \
+{                                                         \
+    return (prev[av_clip(x + k, 0, end)] +                \
+            next[av_clip(x - k, 0, end)] + 1) >> 1;       \
+}
+
+MIDL(uint8_t, 8)
+MIDL(uint16_t, 16)
+
+#define DIFF(type, ss)                                    \
+static int diff_##ss(const type *prev, const type *next,  \
+                     int end, int x, int k, int j)        \
+{                                                         \
+    return abs(prev[av_clip(x + k + j, 0, end)] -         \
+               next[av_clip(x - k + j, 0, end)]);         \
+}
+
+DIFF(uint8_t, 8)
+DIFF(uint16_t, 16)
+
+#define COST(type, ss)                                    \
+static int cost_##ss(const type *prev, const type *next,  \
+                     int end, int x, int k)               \
+{                                                         \
+    const int m = midl_##ss(prev, next, end, x, k);       \
+    const int p = prev[x];                                \
+    const int n = next[x];                                \
+                                                          \
+    return abs(p - m) + abs(n - m);                       \
+}
+
+COST(uint8_t, 8)
+COST(uint16_t, 16)
+
+#define INTERPOLATE(type, ss)                                                  \
+static void interpolate_##ss(uint8_t *ddst,                                    \
+                             const uint8_t *pprev_line,                        \
+                             const uint8_t *nnext_line,                        \
+                             const uint8_t *pprev2_line,                       \
+                             const uint8_t *nnext2_line,                       \
+                             int x, int width, int rslope,                     \
+                             int redge,                                        \
+                             int *K)                                           \
+{                                                                              \
+    type *dst = (type *)ddst;                                                  \
+    const type *prev_line = (const type *)pprev_line;                          \
+    const type *prev2_line = (const type *)pprev2_line;                        \
+    const type *next_line = (const type *)nnext_line;                          \
+    const type *next2_line = (const type *)nnext2_line;                        \
+    const int end = width - 1;                                                 \
+    int k = *K;                                                                \
+    int sd[S], sD[S], dmin = INT_MAX, di = 0;                                  \
+                                                                               \
+    for (int i = -rslope; i <= rslope && abs(k) > rslope; i++) {               \
+        int sum = 0;                                                           \
+                                                                               \
+        for (int j = -redge; j <= redge; j++) {                                \
+            sum += diff_##ss(prev_line,  next_line,  end, x, i, j) +           \
+                   diff_##ss(prev2_line, prev_line,  end, x, i, j) +           \
+                   diff_##ss(next_line,  next2_line, end, x, i, j);            \
+        }                                                                      \
+                                                                               \
+        sD[i + rslope] =   1 * sum +                                           \
+                           4 * cost_##ss(prev_line,  next_line,  end, x, i) +  \
+                         128 * abs(i);                                         \
+                                                                               \
+        dmin = FFMIN(sD[i + rslope], dmin);                                    \
+    }                                                                          \
+                                                                               \
+    for (int i = -rslope; i <= rslope; i++) {                                  \
+        int sum = 0;                                                           \
+                                                                               \
+        for (int j = -redge; j <= redge; j++) {                                \
+            sum += diff_##ss(prev_line,  next_line,  end, x, k + i, j) +       \
+                   diff_##ss(prev2_line, prev_line,  end, x, k + i, j) +       \
+                   diff_##ss(next_line,  next2_line, end, x, k + i, j);        \
+        }                                                                      \
+                                                                               \
+        sd[i + rslope] =   1 * sum +                                           \
+                           4 * cost_##ss(prev_line, next_line, end, x, k + i) +\
+                         128 * abs(k + i);                                     \
+                                                                               \
+        dmin = FFMIN(sd[i + rslope], dmin);                                    \
+    }                                                                          \
+                                                                               \
+    for (int i = -rslope; i <= rslope; i++) {                                  \
+        if (abs(k) > rslope) {                                                 \
+            if (dmin == sD[i + rslope]) {                                      \
+                di = -k + i;                                                   \
+                break;                                                         \
+            }                                                                  \
+        }                                                                      \
+        if (dmin == sd[i + rslope]) {                                          \
+            di = i;                                                            \
+            break;                                                             \
+        }                                                                      \
+    }                                                                          \
+                                                                               \
+    k += di;                                                                   \
+                                                                               \
+    dst[x] = midl_##ss(prev_line, next_line, end, x, k);                       \
+                                                                               \
+    *K = k;                                                                    \
+}
+
+INTERPOLATE(uint8_t, 8)
+INTERPOLATE(uint16_t, 16)
+
+static int deinterlace_slice(AVFilterContext *ctx, void *arg,
+                             int jobnr, int nb_jobs)
+{
+    ESTDIFContext *s = ctx->priv;
+    ThreadData *td = arg;
+    AVFrame *out = td->out;
+    AVFrame *in = td->in;
+    const int rslope = s->rslope;
+    const int redge = s->redge;
+    const int interlaced = in->interlaced_frame;
+    const int tff = (s->field == (s->parity == -1 ?
+                                  in->top_field_first == interlaced :
+                                  s->parity == 0 ? !interlaced : interlaced));
+
+    for (int plane = 0; plane < s->nb_planes; plane++) {
+        const uint8_t *src_data = in->data[plane];
+        uint8_t *dst_data = out->data[plane];
+        const int linesize = s->linesize[plane];
+        const int width = s->planewidth[plane];
+        const int height = s->planeheight[plane];
+        const int src_linesize = in->linesize[plane];
+        const int dst_linesize = out->linesize[plane];
+        const int start = (height * jobnr) / nb_jobs;
+        const int end = (height * (jobnr+1)) / nb_jobs;
+        const uint8_t *prev_line, *prev2_line, *next_line, *next2_line, *in_line;
+        uint8_t *out_line;
+        int y_out;
+
+        y_out = start + (tff ^ (start & 1));
+
+        in_line  = src_data + (y_out * src_linesize);
+        out_line = dst_data + (y_out * dst_linesize);
+
+        while (y_out < end) {
+            memcpy(out_line, in_line, linesize);
+            y_out += 2;
+            in_line  += src_linesize * 2;
+            out_line += dst_linesize * 2;
+        }
+
+        y_out = start + ((!tff) ^ (start & 1));
+        out_line = dst_data + (y_out * dst_linesize);
+
+        for (int y = y_out; y < end; y += 2) {
+            int y_prev2_in = y - 3;
+            int y_next2_in = y + 3;
+            int y_prev_in = y - 1;
+            int y_next_in = y + 1;
+            int k;
+
+            while (y_prev2_in < 0)
+                y_prev2_in += 2;
+
+            while (y_next2_in >= height)
+                y_next2_in -= 2;
+
+            while (y_prev_in < 0)
+                y_prev_in += 2;
+
+            while (y_next_in >= height)
+                y_next_in -= 2;
+
+            prev2_line = src_data + (y_prev2_in * src_linesize);
+            next2_line = src_data + (y_next2_in * src_linesize);
+
+            prev_line = src_data + (y_prev_in * src_linesize);
+            next_line = src_data + (y_next_in * src_linesize);
+
+            k = 0;
+
+            for (int x = 0; x < width; x++) {
+                s->interpolate(out_line,
+                               prev_line, next_line,
+                               prev2_line, next2_line,
+                               x, width, rslope, redge, &k);
+            }
+
+            out_line += 2 * dst_linesize;
+        }
+    }
+
+    return 0;
+}
+
+static int filter(AVFilterContext *ctx, int is_second, AVFrame *in)
+{
+    ESTDIFContext *s = ctx->priv;
+    AVFilterLink *outlink = ctx->outputs[0];
+    AVFrame *out;
+    ThreadData td;
+
+    out = ff_get_video_buffer(outlink, outlink->w, outlink->h);
+    if (!out)
+        return AVERROR(ENOMEM);
+    av_frame_copy_props(out, in);
+    out->interlaced_frame = 0;
+    out->pts = s->pts;
+
+    td.out = out; td.in = in;
+    ctx->internal->execute(ctx, deinterlace_slice, &td, NULL,
+                           FFMIN(s->planeheight[1] / 2, s->nb_threads));
+
+    if (s->mode)
+        s->field = !s->field;
+
+    return ff_filter_frame(outlink, out);
+}
+
+static int config_input(AVFilterLink *inlink)
+{
+    AVFilterContext *ctx = inlink->dst;
+    ESTDIFContext *s = ctx->priv;
+    const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(inlink->format);
+    int ret, depth;
+
+    if ((ret = av_image_fill_linesizes(s->linesize, inlink->format, inlink->w)) < 0)
+        return ret;
+
+    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;
+
+    if (inlink->h < 3) {
+        av_log(ctx, AV_LOG_ERROR, "Video of less than 3 lines is not supported\n");
+        return AVERROR(EINVAL);
+    }
+
+    s->nb_planes = av_pix_fmt_count_planes(inlink->format);
+    s->nb_threads = ff_filter_get_nb_threads(ctx);
+    depth = desc->comp[0].depth;
+    s->interpolate = depth <= 8 ? interpolate_8 : interpolate_16;
+
+    return 0;
+}
+ static int filter_frame(AVFilterLink *inlink, AVFrame *in)
+{
+    AVFilterContext *ctx = inlink->dst;
+    ESTDIFContext *s = ctx->priv;
+    int ret;
+
+    if (!s->prev) {
+        s->prev = in;
+        return 0;
+    }
+
+    if ((s->deint && !in->interlaced_frame) || ctx->is_disabled) {
+        s->prev->pts *= 2;
+        ret = ff_filter_frame(ctx->outputs[0], s->prev);
+        s->prev = in;
+        return ret;
+    }
+
+    s->pts = s->prev->pts * 2;
+    ret = filter(ctx, 0, s->prev);
+    if (ret < 0 || s->mode == 0) {
+        av_frame_free(&s->prev);
+        s->prev = in;
+        return ret;
+    }
+
+    s->pts = s->prev->pts + in->pts;
+    ret = filter(ctx, 1, s->prev);
+    av_frame_free(&s->prev);
+    s->prev = in;
+    return ret;
+}
+
+static int request_frame(AVFilterLink *link)
+{
+    AVFilterContext *ctx = link->src;
+    ESTDIFContext *s = ctx->priv;
+    int ret;
+
+    if (s->eof)
+        return AVERROR_EOF;
+
+    ret = ff_request_frame(ctx->inputs[0]);
+
+    if (ret == AVERROR_EOF && s->prev) {
+        AVFrame *next = av_frame_clone(s->prev);
+
+        if (!next)
+            return AVERROR(ENOMEM);
+
+        next->pts = s->prev->pts + av_rescale_q(1, av_inv_q(ctx->outputs[0]->frame_rate),
+                                                ctx->outputs[0]->time_base);
+        s->eof = 1;
+        ret = filter_frame(ctx->inputs[0], next);
+    } else if (ret < 0) {
+        return ret;
+    }
+
+    return ret;
+}
+
+static av_cold void uninit(AVFilterContext *ctx)
+{
+    ESTDIFContext *s = ctx->priv;
+
+    av_frame_free(&s->prev);
+}
+
+static const AVFilterPad estdif_inputs[] = {
+    {
+        .name          = "default",
+        .type          = AVMEDIA_TYPE_VIDEO,
+        .filter_frame  = filter_frame,
+        .config_props  = config_input,
+    },
+    { NULL }
+};
+
+static const AVFilterPad estdif_outputs[] = {
+    {
+        .name          = "default",
+        .type          = AVMEDIA_TYPE_VIDEO,
+        .config_props  = config_output,
+        .request_frame = request_frame,
+    },
+    { NULL }
+};
+
+AVFilter ff_vf_estdif = {
+    .name          = "estdif",
+    .description   = NULL_IF_CONFIG_SMALL("Apply Edge Slope Tracing deinterlace."),
+    .priv_size     = sizeof(ESTDIFContext),
+    .priv_class    = &estdif_class,
+    .uninit        = uninit,
+    .query_formats = query_formats,
+    .inputs        = estdif_inputs,
+    .outputs       = estdif_outputs,
+    .flags         = AVFILTER_FLAG_SUPPORT_TIMELINE_INTERNAL | AVFILTER_FLAG_SLICE_THREADS,
+    .process_command = ff_filter_process_command,
+};
-- 
2.17.1



More information about the ffmpeg-devel mailing list