[FFmpeg-devel] [PATCH 2/4] avcodec/aom_film_grain: implement AFGS1

Niklas Haas ffmpeg at haasn.xyz
Mon Feb 26 16:35:12 EET 2024


From: Niklas Haas <git at haasn.dev>

Based on the AOMedia Film Grain Synthesis 1 (AFGS1) spec:
  https://aomediacodec.github.io/afgs1-spec/

The parsing has been changed substantially relative to the AV1 film
grain OBU. In particular:

1. There is the possibility of maintaining multiple independent film
   grain parameter sets, and decoders are recommended to pick the one
   most appropriate for the intended display resolution. This is to
   support scalable / multi-level codecs, although this could also be
   used to e.g. switch between different grain profiles without having
   to re-signal the appropriate coefficients.

2. Supporting this, it's possible to *predict* the grain coefficients
   from previously signalled parameter sets, transmitting only the
   residual.

3. When not predicting, the parameter sets are now stored as a series of
   increments, rather than being directly transmitted.

I placed this parser in its own file, rather than h2645_sei.c, since
nothing in the generic AFGS1 film grain payload is specific to T.35.

Note: Due to an ambiguity/mistake in the specification, the parsing of
      AR coefficients is possibly incorrect. See for details:

      https://github.com/AOMediaCodec/afgs1-spec/issues/115
---
 libavcodec/aom_film_grain.c | 227 ++++++++++++++++++++++++++++++++++++
 libavcodec/aom_film_grain.h |  25 ++++
 2 files changed, 252 insertions(+)

diff --git a/libavcodec/aom_film_grain.c b/libavcodec/aom_film_grain.c
index 1bdaafc39f7..5614d3e8d16 100644
--- a/libavcodec/aom_film_grain.c
+++ b/libavcodec/aom_film_grain.c
@@ -29,6 +29,7 @@
 #include "libavutil/imgutils.h"
 
 #include "aom_film_grain.h"
+#include "get_bits.h"
 
 // Common/shared helpers (not dependent on BITDEPTH)
 static inline int get_random_number(const int bits, unsigned *const state) {
@@ -96,6 +97,232 @@ int ff_aom_apply_film_grain(AVFrame *out, const AVFrame *in,
     return AVERROR_INVALIDDATA;
 }
 
+int ff_aom_parse_film_grain_sets(AVFilmGrainAOMParamSets *s,
+                                 const uint8_t *payload, int payload_size)
+{
+    GetBitContext gbc, *gb = &gbc;
+    AVFilmGrainAOMParams *aom;
+    AVFilmGrainAOMParamSet *fgps, *ref = NULL;
+    int ret, num_sets, n, i, uv, num_y_coeffs, update_grain, luma_only;
+
+    ret = init_get_bits8(gb, payload, payload_size);
+    if (ret < 0)
+        return ret;
+
+    s->enable = get_bits1(gb);
+    if (!s->enable)
+        return 0;
+
+    skip_bits(gb, 4); // reserved
+    num_sets = get_bits(gb, 3);
+    for (n = 0; n < num_sets; n++) {
+        int payload_4byte, payload_size, set_idx, apply_units_log2, vsc_flag;
+        int predict_scaling, predict_y_scaling, predict_uv_scaling[2];
+        int payload_bits, start_position;
+
+        start_position = get_bits_count(gb);
+        payload_4byte = get_bits1(gb);
+        payload_size = get_bits(gb, payload_4byte ? 2 : 8);
+        set_idx = get_bits(gb, 3);
+        fgps = &s->sets[set_idx];
+
+        fgps->apply_grain = get_bits1(gb);
+        if (!fgps->apply_grain)
+            continue;
+
+        fgps->grain_seed = get_bits(gb, 16);
+        update_grain = get_bits1(gb);
+        if (!update_grain)
+            continue;
+
+        apply_units_log2  = get_bits(gb, 4);
+        fgps->apply_width  = get_bits(gb, 12) << apply_units_log2;
+        fgps->apply_height = get_bits(gb, 12) << apply_units_log2;
+        luma_only = get_bits1(gb);
+        if (luma_only) {
+            fgps->subx = fgps->suby = 0;
+        } else {
+            fgps->subx = get_bits1(gb);
+            fgps->suby = get_bits1(gb);
+        }
+
+        vsc_flag = get_bits1(gb); // video_signal_characteristics_flag
+        if (vsc_flag) {
+            int cicp_flag;
+            skip_bits(gb, 3); // bit_depth_minus8
+            cicp_flag = get_bits1(gb);
+            if (cicp_flag)
+                skip_bits(gb, 8 + 8 + 8 + 1); // cicp_info
+        }
+
+        aom = &fgps->params;
+        predict_scaling = get_bits1(gb);
+        if (predict_scaling && (!ref || ref == fgps))
+            goto error; // prediction must be from valid, different set
+
+        predict_y_scaling = predict_scaling ? get_bits1(gb) : 0;
+        if (predict_y_scaling) {
+            int y_scale, y_offset, bits_res;
+            y_scale = get_bits(gb, 9) - 256;
+            y_offset = get_bits(gb, 9) - 256;
+            bits_res = get_bits(gb, 3);
+            if (bits_res) {
+                int res[14], pred, granularity;
+                aom->num_y_points = ref->params.num_y_points;
+                for (i = 0; i < aom->num_y_points; i++)
+                    res[i] = get_bits(gb, bits_res);
+                granularity = get_bits(gb, 3);
+                for (i = 0; i < aom->num_y_points; i++) {
+                    pred = ref->params.y_points[i][1];
+                    pred = ((pred * y_scale + 8) >> 4) + y_offset;
+                    pred += (res[i] - (1 << (bits_res - 1))) * granularity;
+                    aom->y_points[i][0] = ref->params.y_points[i][0];
+                    aom->y_points[i][1] = av_clip_uint8(pred);
+                }
+            }
+        } else {
+            aom->num_y_points = get_bits(gb, 4);
+            if (aom->num_y_points > 14) {
+                goto error;
+            } else if (aom->num_y_points) {
+                int bits_inc, bits_scaling;
+                int y_value = 0;
+                bits_inc = get_bits(gb, 3) + 1;
+                bits_scaling = get_bits(gb, 2) + 5;
+                for (i = 0; i < aom->num_y_points; i++) {
+                    y_value += get_bits(gb, bits_inc);
+                    if (y_value > UINT8_MAX)
+                        goto error;
+                    aom->y_points[i][0] = y_value;
+                    aom->y_points[i][1] = get_bits(gb, bits_scaling);
+                }
+            }
+        }
+
+        if (luma_only) {
+            aom->chroma_scaling_from_luma = 0;
+            aom->num_uv_points[0] = aom->num_uv_points[1] = 0;
+        } else {
+            aom->chroma_scaling_from_luma = get_bits1(gb);
+            if (aom->chroma_scaling_from_luma) {
+                aom->num_uv_points[0] = aom->num_uv_points[1] = 0;
+            } else {
+                for (uv = 0; uv < 2; uv++) {
+                    predict_uv_scaling[uv] = predict_scaling ? get_bits1(gb) : 0;
+                    if (predict_uv_scaling[uv]) {
+                        int uv_scale, uv_offset, bits_res;
+                        uv_scale = get_bits(gb, 9) - 256;
+                        uv_offset = get_bits(gb, 9) - 256;
+                        bits_res = get_bits(gb, 3);
+                        aom->uv_mult[uv] = ref->params.uv_mult[uv];
+                        aom->uv_mult_luma[uv] = ref->params.uv_mult_luma[uv];
+                        aom->uv_offset[uv] = ref->params.uv_offset[uv];
+                        if (bits_res) {
+                            int res[10], pred, granularity;
+                            aom->num_uv_points[uv] = ref->params.num_uv_points[uv];
+                            for (i = 0; i < aom->num_uv_points[uv]; i++)
+                                res[i] = get_bits(gb, bits_res);
+                            granularity = get_bits(gb, 3);
+                            for (i = 0; i < aom->num_uv_points[uv]; i++) {
+                                pred = ref->params.uv_points[uv][i][1];
+                                pred = ((pred * uv_scale + 8) >> 4) + uv_offset;
+                                pred += (res[i] - (1 << (bits_res - 1))) * granularity;
+                                aom->uv_points[uv][i][0] = ref->params.uv_points[uv][i][0];
+                                aom->uv_points[uv][i][1] = av_clip_uint8(pred);
+                            }
+                        }
+                    } else {
+                        int bits_inc, bits_scaling, uv_offset;
+                        int uv_value = 0;
+                        aom->num_uv_points[uv] = get_bits(gb, 4);
+                        if (aom->num_uv_points[uv] > 10)
+                            goto error;
+                        bits_inc = get_bits(gb, 3) + 1;
+                        bits_scaling = get_bits(gb, 2) + 5;
+                        uv_offset = get_bits(gb, 8);
+                        for (i = 0; i < aom->num_uv_points[uv]; i++) {
+                            uv_value += get_bits(gb, bits_inc);
+                            if (uv_value > UINT8_MAX)
+                                goto error;
+                            aom->uv_points[uv][i][0] = uv_value;
+                            aom->uv_points[uv][i][1] = get_bits(gb, bits_scaling) + uv_offset;
+                        }
+                    }
+                }
+            }
+        }
+
+        aom->scaling_shift = get_bits(gb, 2) + 8;
+        aom->ar_coeff_lag = get_bits(gb, 2);
+        num_y_coeffs = 2 * aom->ar_coeff_lag * (aom->ar_coeff_lag + 1);
+        if (aom->num_y_points) {
+            int ar_bits = get_bits(gb, 2) + 5;
+            for (i = 0; i < num_y_coeffs; i++)
+                aom->ar_coeffs_y[i] = get_bits(gb, ar_bits) - (1 << (ar_bits - 1));
+        }
+        for (uv = 0; uv < 2; uv++) {
+            if (aom->chroma_scaling_from_luma || aom->num_uv_points[uv]) {
+                int ar_bits = get_bits(gb, 2) + 5;
+                for (i = 0; i < num_y_coeffs + !!aom->num_y_points; i++)
+                    aom->ar_coeffs_uv[uv][i] = get_bits(gb, ar_bits) - (1 << (ar_bits - 1));
+            }
+        }
+        aom->ar_coeff_shift = get_bits(gb, 2) + 6;
+        aom->grain_scale_shift = get_bits(gb, 2);
+        for (uv = 0; uv < 2; uv++) {
+            if (aom->num_uv_points[uv] && !predict_uv_scaling[uv]) {
+                aom->uv_mult[uv]      = get_bits(gb, 8) - 128;
+                aom->uv_mult_luma[uv] = get_bits(gb, 8) - 128;
+                aom->uv_offset[uv]    = get_bits(gb, 9) - 256;
+            }
+        }
+        aom->overlap_flag = get_bits1(gb);
+        aom->limit_output_range = get_bits1(gb);
+
+        // use first set as reference only if it was fully transmitted
+        if (n == 0)
+            ref = fgps;
+
+        payload_bits = get_bits_count(gb) - start_position;
+        if (payload_bits > payload_size * 8)
+            goto error;
+        skip_bits(gb, payload_size * 8 - payload_bits);
+    }
+    return 0;
+
+error:
+    s->enable = 0;
+    return AVERROR_INVALIDDATA;
+}
+
+const AVFilmGrainAOMParamSet *ff_aom_select_film_grain_set(const AVFilmGrainAOMParamSets *s,
+                                                           const AVFrame *frame)
+{
+    const AVFilmGrainAOMParamSet *fgps, *best;
+    const AVPixFmtDescriptor *desc = av_pix_fmt_desc_get(frame->format);
+    if (!s->enable || !desc)
+        return NULL;
+
+    best = NULL;
+    for (int i = 0; i < 8; i++) {
+        fgps = &s->sets[i];
+        if (!fgps->apply_width  ||
+            !fgps->apply_height ||
+            fgps->apply_width  > frame->width  ||
+            fgps->apply_height > frame->height ||
+            fgps->subx != desc->log2_chroma_w ||
+            fgps->suby != desc->log2_chroma_h)
+            continue;
+
+        if (!best ||
+            fgps->apply_width  > best->apply_width ||
+            fgps->apply_height > best->apply_height)
+            best = fgps;
+    }
+
+    return best;
+}
+
 // Taken from the AV1 spec. Range is [-2048, 2047], mean is 0 and stddev is 512
 static const int16_t gaussian_sequence[2048] = {
     56,    568,   -180,  172,   124,   -84,   172,   -64,   -900,  24,   820,
diff --git a/libavcodec/aom_film_grain.h b/libavcodec/aom_film_grain.h
index 5d772bd7d17..b985451dbc3 100644
--- a/libavcodec/aom_film_grain.h
+++ b/libavcodec/aom_film_grain.h
@@ -30,9 +30,34 @@
 
 #include "libavutil/film_grain_params.h"
 
+// Stand-alone AFGS1 metadata parameter set
+typedef struct AVFilmGrainAOMParamSet {
+    int apply_grain;
+    int apply_width;
+    int apply_height;
+    int subx, suby;
+    uint16_t grain_seed;
+    AVFilmGrainAOMParams params;
+} AVFilmGrainAOMParamSet;
+
+typedef struct AVFilmGrainAOMParamSets {
+    int enable;
+    AVFilmGrainAOMParamSet sets[8];
+} AVFilmGrainAOMParamSets;
+
 // Synthesizes film grain on top of `in` and stores the result to `out`. `out`
 // must already have been allocated and set to the same size and format as `in`.
 int ff_aom_apply_film_grain(AVFrame *out, const AVFrame *in,
                             const AVFilmGrainParams *params);
 
+// Parse AFGS1 parameter sets from an ITU-T T.35 payload. Returns 0 on success,
+// or a negative error code.
+int ff_aom_parse_film_grain_sets(AVFilmGrainAOMParamSets *s,
+                                 const uint8_t *payload, int payload_size);
+
+// Select the most appropriate film grain parameter set for a given
+// frame. Returns the parameter set, or NULL if none was selected.
+const AVFilmGrainAOMParamSet *ff_aom_select_film_grain_set(const AVFilmGrainAOMParamSets *s,
+                                                           const AVFrame *frame);
+
 #endif /* AVCODEC_AOM_FILM_GRAIN_H */
-- 
2.43.2



More information about the ffmpeg-devel mailing list