Browse Source

AAC: Fix AAC prediction when used in conjunction with the CPE

Patch by Alex Converse (alex converse gmail com)

Originally committed as revision 16250 to svn://svn.ffmpeg.org/ffmpeg/trunk
tags/v0.5
Alex Converse Robert Swain 16 years ago
parent
commit
aab5413323
2 changed files with 15 additions and 9 deletions
  1. +14
    -8
      libavcodec/aac.c
  2. +1
    -1
      libavcodec/aac.h

+ 14
- 8
libavcodec/aac.c View File

@@ -753,7 +753,7 @@ static int decode_spectrum_and_dequant(AACContext * ac, float coef[1024], GetBit
const int dim = cur_band_type >= FIRST_PAIR_BT ? 2 : 4; const int dim = cur_band_type >= FIRST_PAIR_BT ? 2 : 4;
const int is_cb_unsigned = IS_CODEBOOK_UNSIGNED(cur_band_type); const int is_cb_unsigned = IS_CODEBOOK_UNSIGNED(cur_band_type);
int group; int group;
if (cur_band_type == ZERO_BT) {
if (cur_band_type == ZERO_BT || cur_band_type == INTENSITY_BT2 || cur_band_type == INTENSITY_BT) {
for (group = 0; group < ics->group_len[g]; group++) { for (group = 0; group < ics->group_len[g]; group++) {
memset(coef + group * 128 + offsets[i], 0, (offsets[i+1] - offsets[i])*sizeof(float)); memset(coef + group * 128 + offsets[i], 0, (offsets[i+1] - offsets[i])*sizeof(float));
} }
@@ -771,7 +771,7 @@ static int decode_spectrum_and_dequant(AACContext * ac, float coef[1024], GetBit
coef[group*128+k] *= scale; coef[group*128+k] *= scale;
} }
} }
}else if (cur_band_type != INTENSITY_BT2 && cur_band_type != INTENSITY_BT) {
}else {
for (group = 0; group < ics->group_len[g]; group++) { for (group = 0; group < ics->group_len[g]; group++) {
for (k = offsets[i]; k < offsets[i+1]; k += dim) { for (k = offsets[i]; k < offsets[i+1]; k += dim) {
const int index = get_vlc2(gb, vlc_spectral[cur_band_type - 1].table, 6, 3); const int index = get_vlc2(gb, vlc_spectral[cur_band_type - 1].table, 6, 3);
@@ -911,21 +911,21 @@ static void apply_prediction(AACContext * ac, SingleChannelElement * sce) {
int sfb, k; int sfb, k;


if (!sce->ics.predictor_initialized) { if (!sce->ics.predictor_initialized) {
reset_all_predictors(sce->ics.predictor_state);
reset_all_predictors(sce->predictor_state);
sce->ics.predictor_initialized = 1; sce->ics.predictor_initialized = 1;
} }


if (sce->ics.window_sequence[0] != EIGHT_SHORT_SEQUENCE) { if (sce->ics.window_sequence[0] != EIGHT_SHORT_SEQUENCE) {
for (sfb = 0; sfb < ff_aac_pred_sfb_max[ac->m4ac.sampling_index]; sfb++) { for (sfb = 0; sfb < ff_aac_pred_sfb_max[ac->m4ac.sampling_index]; sfb++) {
for (k = sce->ics.swb_offset[sfb]; k < sce->ics.swb_offset[sfb + 1]; k++) { for (k = sce->ics.swb_offset[sfb]; k < sce->ics.swb_offset[sfb + 1]; k++) {
predict(ac, &sce->ics.predictor_state[k], &sce->coeffs[k],
predict(ac, &sce->predictor_state[k], &sce->coeffs[k],
sce->ics.predictor_present && sce->ics.prediction_used[sfb]); sce->ics.predictor_present && sce->ics.prediction_used[sfb]);
} }
} }
if (sce->ics.predictor_reset_group) if (sce->ics.predictor_reset_group)
reset_predictor_group(sce->ics.predictor_state, sce->ics.predictor_reset_group);
reset_predictor_group(sce->predictor_state, sce->ics.predictor_reset_group);
} else } else
reset_all_predictors(sce->ics.predictor_state);
reset_all_predictors(sce->predictor_state);
} }


/** /**
@@ -983,7 +983,7 @@ static int decode_ics(AACContext * ac, SingleChannelElement * sce, GetBitContext
if (decode_spectrum_and_dequant(ac, out, gb, sce->sf, pulse_present, &pulse, ics, sce->band_type) < 0) if (decode_spectrum_and_dequant(ac, out, gb, sce->sf, pulse_present, &pulse, ics, sce->band_type) < 0)
return -1; return -1;


if(ac->m4ac.object_type == AOT_AAC_MAIN)
if(ac->m4ac.object_type == AOT_AAC_MAIN && !common_window)
apply_prediction(ac, sce); apply_prediction(ac, sce);


return 0; return 0;
@@ -1086,8 +1086,14 @@ static int decode_cpe(AACContext * ac, GetBitContext * gb, int elem_id) {
if ((ret = decode_ics(ac, &cpe->ch[1], gb, common_window, 0))) if ((ret = decode_ics(ac, &cpe->ch[1], gb, common_window, 0)))
return ret; return ret;


if (common_window && ms_present)
if (common_window) {
if (ms_present)
apply_mid_side_stereo(cpe); apply_mid_side_stereo(cpe);
if (ac->m4ac.object_type == AOT_AAC_MAIN) {
apply_prediction(ac, &cpe->ch[0]);
apply_prediction(ac, &cpe->ch[1]);
}
}


apply_intensity_stereo(cpe, ms_present); apply_intensity_stereo(cpe, ms_present);
return 0; return 0;


+ 1
- 1
libavcodec/aac.h View File

@@ -163,7 +163,6 @@ typedef struct {
int predictor_initialized; int predictor_initialized;
int predictor_reset_group; int predictor_reset_group;
uint8_t prediction_used[41]; uint8_t prediction_used[41];
PredictorState predictor_state[MAX_PREDICTORS];
} IndividualChannelStream; } IndividualChannelStream;


/** /**
@@ -226,6 +225,7 @@ typedef struct {
DECLARE_ALIGNED_16(float, coeffs[1024]); ///< coefficients for IMDCT DECLARE_ALIGNED_16(float, coeffs[1024]); ///< coefficients for IMDCT
DECLARE_ALIGNED_16(float, saved[512]); ///< overlap DECLARE_ALIGNED_16(float, saved[512]); ///< overlap
DECLARE_ALIGNED_16(float, ret[1024]); ///< PCM output DECLARE_ALIGNED_16(float, ret[1024]); ///< PCM output
PredictorState predictor_state[MAX_PREDICTORS];
} SingleChannelElement; } SingleChannelElement;


/** /**


Loading…
Cancel
Save