Calculate wavelet energy for perceptual deltaq mode
The stats based on wavelet energy calculated in first
pass is required in case of perceptual deltaq mode. Thus,
wavelet energy calculation is enabled for those cases.
Change-Id: Ifbffd7e1fc8dc7da87aa8b9e4db12603207c1488
diff --git a/av1/encoder/encoder.c b/av1/encoder/encoder.c
index c29b226..4ce34f0 100644
--- a/av1/encoder/encoder.c
+++ b/av1/encoder/encoder.c
@@ -3079,6 +3079,42 @@
return err;
}
+#if !CONFIG_REALTIME_ONLY
+static void calculate_frame_avg_haar_energy(AV1_COMP *cpi) {
+ TWO_PASS *const twopass = &cpi->ppi->twopass;
+ const FIRSTPASS_STATS *const total_stats =
+ twopass->stats_buf_ctx->total_stats;
+
+ if (is_one_pass_rt_params(cpi) ||
+ (cpi->oxcf.q_cfg.deltaq_mode != DELTA_Q_PERCEPTUAL) ||
+ (is_fp_wavelet_energy_invalid(total_stats) == 0))
+ return;
+
+ const int num_mbs = (cpi->oxcf.resize_cfg.resize_mode != RESIZE_NONE)
+ ? cpi->initial_mbs
+ : cpi->common.mi_params.MBs;
+ const YV12_BUFFER_CONFIG *const unfiltered_source = cpi->unfiltered_source;
+ const uint8_t *const src = unfiltered_source->y_buffer;
+ const int hbd = unfiltered_source->flags & YV12_FLAG_HIGHBITDEPTH;
+ const int stride = unfiltered_source->y_stride;
+ const BLOCK_SIZE fp_block_size =
+ get_fp_block_size(cpi->is_screen_content_type);
+ const int fp_block_size_width = block_size_wide[fp_block_size];
+ const int fp_block_size_height = block_size_high[fp_block_size];
+ const int num_unit_cols =
+ get_num_blocks(unfiltered_source->y_crop_width, fp_block_size_width);
+ const int num_unit_rows =
+ get_num_blocks(unfiltered_source->y_crop_height, fp_block_size_height);
+ const int num_8x8_cols = num_unit_cols * (fp_block_size_width / 8);
+ const int num_8x8_rows = num_unit_rows * (fp_block_size_height / 8);
+ int64_t frame_avg_wavelet_energy = av1_haar_ac_sad_mxn_uint8_input(
+ src, stride, hbd, num_8x8_rows, num_8x8_cols);
+
+ twopass->frame_avg_haar_energy =
+ log(((double)frame_avg_wavelet_energy / num_mbs) + 1.0);
+}
+#endif
+
// Process the wiener variance in 16x16 block basis.
static int qsort_comp(const void *elem1, const void *elem2) {
int a = *((const int *)elem1);
@@ -3355,6 +3391,10 @@
av1_set_screen_content_options(cpi, features);
}
+#if !CONFIG_REALTIME_ONLY
+ calculate_frame_avg_haar_energy(cpi);
+#endif
+
// frame type has been decided outside of this function call
cm->cur_frame->frame_type = current_frame->frame_type;
diff --git a/av1/encoder/firstpass.c b/av1/encoder/firstpass.c
index 63e4dc0..00203aa 100644
--- a/av1/encoder/firstpass.c
+++ b/av1/encoder/firstpass.c
@@ -357,6 +357,9 @@
return raw_err_stdev;
}
+static AOM_INLINE int calc_wavelet_energy(const AV1EncoderConfig *oxcf) {
+ return oxcf->q_cfg.deltaq_mode == DELTA_Q_PERCEPTUAL;
+}
typedef struct intra_pred_block_pass1_args {
const SequenceHeader *seq_params;
MACROBLOCK *x;
@@ -552,6 +555,24 @@
// Accumulate the intra error.
stats->intra_error += (int64_t)this_intra_error;
+ // Stats based on wavelet energy is used in the following cases :
+ // 1. ML model which predicts if a flat structure (golden-frame only structure
+ // without ALT-REF and Internal-ARFs) is better. This ML model is enabled in
+ // constant quality mode under certain conditions.
+ // 2. Delta qindex mode is set as DELTA_Q_PERCEPTUAL.
+ // Thus, wavelet energy calculation is enabled for the above cases.
+ if (calc_wavelet_energy(&cpi->oxcf)) {
+ const int hbd = is_cur_buf_hbd(xd);
+ const int stride = x->plane[0].src.stride;
+ const int num_8x8_rows = block_size_high[fp_block_size] / 8;
+ const int num_8x8_cols = block_size_wide[fp_block_size] / 8;
+ const uint8_t *buf = x->plane[0].src.buf;
+ stats->frame_avg_wavelet_energy += av1_haar_ac_sad_mxn_uint8_input(
+ buf, stride, hbd, num_8x8_rows, num_8x8_cols);
+ } else {
+ stats->frame_avg_wavelet_energy = INVALID_FP_STATS_TO_PREDICT_FLAT_GOP;
+ }
+
return this_intra_error;
}
diff --git a/av1/encoder/firstpass.h b/av1/encoder/firstpass.h
index 551e604..31ee7fc 100644
--- a/av1/encoder/firstpass.h
+++ b/av1/encoder/firstpass.h
@@ -354,6 +354,11 @@
struct AV1EncoderConfig;
struct TileDataEnc;
+static INLINE int is_fp_wavelet_energy_invalid(
+ const FIRSTPASS_STATS *fp_stats) {
+ return (fp_stats->frame_avg_wavelet_energy < 0);
+}
+
static INLINE BLOCK_SIZE get_fp_block_size(int is_screen_content_type) {
return (is_screen_content_type ? BLOCK_8X8 : BLOCK_16X16);
}
diff --git a/av1/encoder/pass2_strategy.c b/av1/encoder/pass2_strategy.c
index 80f165f..a43d816 100644
--- a/av1/encoder/pass2_strategy.c
+++ b/av1/encoder/pass2_strategy.c
@@ -3359,6 +3359,13 @@
// applied when combining MB error values for the frame.
twopass->mb_av_energy = log((this_frame_ptr->intra_error) + 1.0);
+ const FIRSTPASS_STATS *const total_stats =
+ twopass->stats_buf_ctx->total_stats;
+ if (is_fp_wavelet_energy_invalid(total_stats) == 0) {
+ twopass->frame_avg_haar_energy =
+ log((this_frame_ptr->frame_avg_wavelet_energy) + 1.0);
+ }
+
// Set the frame content type flag.
if (this_frame_ptr->intra_skip_pct >= FC_ANIMATION_THRESH)
twopass->fr_content_type = FC_GRAPHICS_ANIMATION;
